PAGES
Subscribe Get it delivered — five to ten releases we picked each morning from the tools you run, with every release in the issue on the web. Compare plans →

The Cyber Toolchain — issue 146, November 21, 2026

THE CYBER TOOLCHAIN NO. 146
Tail
THE DAILY RELEASE FIREHOSE
PUBLISHED NOVEMBER 21, 2026 · EVERY DAY
EDITIONS tail grep head ls diff uniq

The daily firehose — everything the toolchain shipped today, already filtered.

We read every release from the 1326 tools on our watchlist at the source — GitHub and GitLab release notes, vendor release pages and changelogs, project blogs and feeds, vendor press releases, and the source code behind the tag. Bug-fix-only releases and non-product newsroom noise are dropped; what's left is summarized down to the new capability, how to try it, and any screenshots or videos the release itself published. Every entry links to the sources it was built from.

VIEW
ISSUE VIEW scroll
INFORMATION DENSITY everything
ISSUE WIDTH reading width
Do you prefer scroll?
ADDITIONAL FILTERS
$ tct list   # 43 tools matched
Same issue, same prompt, two writers:

What stands out across today's releases, grouped by what it lets you do. Every tool named links to its entry below.

Three unrelated things stand out. Qualys VMDR can now mitigate Java4Shell by applying a JVMJVMThe Java Virtual Machine, a runtime maintained by Oracle (and implemented by others) that executes Java bytecode on any OS, giving cyber tools a portable, garbage-collected environment with a large ecosystem of libraries. flag to eligible hosts with one-click rollback, buying time before patching. GitGuardian stops hooks and AI coding agents blocking on secrets it has confirmed dead. And GitHub Enterprise Server 3.23 RC brings AI secret detection on-appliance for teams that cannot send code out.

  • protect

    Neutralize Java4Shell on hosts you cannot patch this week

    Qualys applies a JVM flag to eligible hosts and reports process-level coverage, so you can see which running JVMs are actually covered rather than assuming, and roll back in one click if an app misbehaves. GitHub's 3.23 RC pairs a Dependabot cooldown with wider malware advisory coverage, holding back freshly published package versions long enough for a compromised release to surface.

    Qualys VMDR · GitHub

  • protect

    Stop burning developer time on secrets and findings that are already dead

    GitGuardian filters confirmed-invalid secrets out of results, which matters most for pre-commitpre-commitAn open-source framework for managing and running git hooks, maintained by the pre-commit project, that lets cyber tools enforce checks automatically on every commit before code reaches a repository. hooks and AI coding agents that halt on every match regardless of whether the credential still works. Socket extends reachability to Java/Android so an unexercised dependency CVE stops competing with ones your code actually calls, and Sysdig's Stage filter separates registry, pipeline and runtime findings so a runtime-only exposure is not buried in base-image noise.

    GitGuardian SaaS · Socket · Sysdig Secure

  • govern

    Run an access certification campaign with evidence an auditor will accept

    JumpCloud's multi-stage approval chains and signed exports replace the spreadsheet-and-email access review most teams still run quarterly, and the signature is what turns the result into evidence rather than an assertion. CISOCISOChief Information Security Officer, the executive responsible for an organization's information and cyber risk strategy; cyber tools often integrate reporting or compliance features aimed at this audience. Assistant adds findings-to-assessment linking so a control gap traces back to the specific finding that opened it.

    JumpCloud

Does Opus 5 read better?
Offensive Security
◆  AI-Powered Offensive Security

dark-moon

Sources Release notes → v1.7.0 Notes

Dark-moon is an open-source autonomous AI penetration testing. 50 specialist agents across web, API, cloud, Active Directory, Kubernetes, CI/CD and AI/LLM (OWASP LLM Top 10).

dark-moon adds Cyber Score v1, a Prioritized Treatment Plan, Terminal Explorer (OpenCode plugin), and attack pathsattack pathsA possible sequence of steps an attacker could take to reach a valuable target. It connects weaknesses, permissions and reachable systems rather than treating each finding separately. with enabled_by agent declarations.

└──▷ GET THIS VERSION
$ git clone --branch v1.7.0 https://github.com/ASCIT31/Dark-Moon.git
# already have the repo? check out this version:
$ git checkout v1.7.0
  • New Cyber Score v1 deterministic rating with coverage gates ('not rated / limited / rated') and a published methodology in docs/cyber-score-methodology.md. Not checked yet
  • New Prioritized Treatment Plan gives one action per asset × weakness class, with indicative effort, status history, retest-based verification, and risk acceptance with a named owner and expiry. Not checked yet
  • New Terminal Explorer (OpenCode plugin) adds a keyboard-driven workspace covering the executive block, projects, campaigns, findings, treatment plan, attack paths, frameworks, profile, and legend — readable at 80 columns. Not checked yet
  • Agents now declare attack-path links using enabled_by to show what enabled what, with every link held as a hypothesis unless evidence supports it. Not checked yet
  • dashboard_* writes are now bound to the calling run when a run key is present, scoping dashboard mutations to their originating run. Not checked yet
+2 more show less
  • Tool-output filter now fails closed, masking private keys, table secrets, and /proc/net hex addresses before the model. Not checked yet
  • Target-safety rule added to every sub-agent, blocking state-changing and self-hanging probes. Not checked yet

7 details in Everything

└──▷ BREAKING ON UPGRADE
  • !A terminal campaign status can no longer be re-finalised after it has been closed.
  • !Agent labels can no longer inflate the Cyber Score (coverage gate enforced server-side).
Was this useful?

Keygraph shannon

Sources Release notes → v3.6.0 Notes Rank Why this rank top features scored 35/35 — 35 How ranking works →

Shannon is an AI pentester for web applications and APIs. It analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production.

Shannon's v3.6.0 release adds a documentation-backed skill plugin for extending agent capabilities and introduces configurable concurrency, timeout, and retry limits for scan workers.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 2 FEATURES 0 OF 2 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Documentation-backed Shannon skill plugin Not checked yet new 35

Shannon adds a documentation-backed Shannon skill plugin to extend agent capabilities with knowledge drawn from your own documentation.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Names the plugin but no config or usage detail given v3.6.0
02 00
Concurrency, timeout and retry limits for scan workers Not checked yet improved 35

Shannon adds concurrency and timeout limits, plus a cap on agent retries, to the scan worker.

— States limits exist but no exact values or flags named v3.6.0
Was this useful?
◆  Hacking Hardware

ESP32 Marauder

Sources Release notes → v1.21.0 Notes Rank Why this rank top features scored 35 — 35 How ranking works →

ESP32 Marauder is a suite of WiFi/Bluetooth offensive and defensive tools for the ESP32.

ESP32 Marauder adds the ability to stream files off the SD card over the serial interface.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 1 FEATURE 0 OF 1 FEATURE most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Framed SD file streaming over serial Not checked yet new 35

Adds framed SD file streaming over serial, enabling file transfer from the SD card via the serial interface.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Names mechanism but no commands or config details given v1.21.0
Was this useful?
◆  Post-Exploitation

proxychains-ng

Sources Release notes → v4.18 Notes Rank Why this rank top features scored 83/65, +6 runnable example — 80 How ranking works →

proxychains-ng v4.18 adds support for routing through a Unix domain socket as the first proxy in a chain and introduces a new configure option for building fat binaries on Apple M2.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

proxychains ng (new generation) - a preloader which hooks calls to sockets in dynamically linked programs and redirects it through one or more socks/http proxies. continuation of the unmaintained proxychains project.

proxychains-ng v4.18 adds support for routing through a Unix domain socket as the first proxy in a chain and introduces a new configure option for building fat binaries on Apple M2.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 2 FEATURES 0 OF 2 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Unix domain socket as first proxy Not checked yet new 83

The first proxy in a chain can now be configured as a Unix domain socket instead of a TCPTCPA network protocol that delivers a reliable, ordered stream of bytes between two endpoints. It detects lost data and retransmits it. endpoint, e.g.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ socks5 unix:/home/user/.ssh/sock/socks.sock in proxychains.conf, letting proxychains reach a SOCKSSOCKSA proxy protocol that relays TCP (and optionally UDP) connections through an intermediary server, letting cyber tools route traffic through a pivot host or anonymizing layer without needing raw network access. proxy bound to a local socket (such as one created by ssh -D with StreamLocalBindMask) rather than an exposed TCP port.

Route a tool through an SSH SOCKS proxy bound to a local socket that only its owner can access, avoiding an exposed TCP port.
$ ssh -N -o StreamLocalBindMask=0177 -D ~/.ssh/sock/socks.sock relay
Point proxychains at the socket-bound SOCKS proxy by adding a socks5 unix: entry as the first (and only) entry in the proxy list.
ini
[ProxyList]
socks5	unix:/home/user/.ssh/sock/socks.sock
— Names exact config syntax and file, with runnable example v4.18
02 00
Fat binary build for Apple M2 Not checked yet new 65

A new --fat-binary-m2 configure option builds a fat binary targeting Apple M2, usable as ./configure --prefix=/usr --sysconfdir=/etc --fat-binary-m2 && make.

Compile a fat binary that runs natively on Apple M2 hardware.
$ ./configure --prefix=/usr --sysconfdir=/etc --fat-binary-m2 && make
— Exact flag and build command given, but no further mechanism detail v4.18
Was this useful?
Application Security
◆  Code Security

psalm-plugin-laravel

Sources Release notes → v4.23.1 Notes Rank Why this rank top features scored 90/90/80, +15 breaking change, +6 runnable example — 107.67 How ranking works →

Psalm-plugin-laravel is a security-focused Laravel static analyser

Psalm-plugin-laravel significantly expanded its taint-tracking coverage this window — flowing taint through Stringable chains, open redirects, raw DB columns, cookies and session values — while also adding Laravel 13 console Signature-attribute support and bumping its minimum Psalm requirement.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 8 FEATURES 0 OF 8 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Taint propagation through Stringable chains Not checked yet improved 90

Taint now propagates through Stringable method chains — trim(), upper(), after(), limit(), append(), and toString() preserve taint, while hash(), encrypt(), toBase64(), and numbers() drop it, and slug() and stripTags() remove the HTML taint specifically.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Catch XSSXSSCross-site scripting, an attack that runs malicious script in another user's browser through a trusted site. It can steal data or perform actions using the user's session. from user input that travels through a Stringable chain before being echoed — previously missed, now reported as TaintedHtml.
php
$name = $request->str('name');
echo $name->trim()->upper();
// TaintedHtml — taint is preserved through the chain
— Names exact methods and behaviour with a worked code example v4.23.1
02 00
Laravel 13 #[Signature] attribute support Not checked yet new 90

Reads console command signatures from Laravel 13's #[Signature] attribute so that argument()/option() type narrowing, hasArgument()/hasOption(), and the InvalidConsoleArgumentName/InvalidConsoleOptionName checks all work for attribute-based commands, mirroring how the attribute overrides protected $signature in Laravel itself.

Annotate a Laravel 13 console command with #[Signature] so Psalm narrows argument and option types accurately without a $signature property.
php
#[Signature('app:demo {name} {--tag=*}')]
final class Demo extends Command {
    public function handle(): void {
        $name = $this->argument('name'); // string
        foreach ($this->option('tag') as $tag) {} // array<int, string>
    }
}
— Full mechanism, named checks, and runnable example v4.22.6
03 00
Open redirect and URL component taint detection Not checked yet new 80

Detects user-controlled redirect targets as TaintedRedirect and injected URL parts as TaintedUrlComponent or TaintedUrlPath on redirect(), Redirector, ResponseFactory and its contract.

Catch an open redirect where user input is appended directly to a URL passed to redirect()
php
redirect('https://example.com/' . (string) $request->input('page'));
// psalm-laravel reports: TaintedUrlComponent
— Names exact taint types and APIs, includes runnable example v4.23.0
04 00
Raw DB column taint tracking Not checked yet new 80

Tracks what raw DB/Connection queries write into each column and reports taint where a subsequent query reads that column back, without needing Eloquent.

Detect taint flowing from a raw INSERT into a later SELECT on the same column, without needing Eloquent
php
DB::insert('INSERT INTO comments (author, body) VALUES (?, ?)', ['admin', $request->input('body')]);

$comment = DB::selectOne('SELECT author, body FROM comments');
echo (string) $comment->body;
// psalm-laravel reports: TaintedHtml
echo (string) $comment->author;
// no issue
— Clear mechanism with before/after code example v4.23.0
05 00
Taint clearance via regex validation Not checked yet new 75

Trusts a value that Validator::make() or validator() checked against a regex rule once the validator passes, clearing TaintedHtml and similar issues.

Clear taint on a string once it has been validated against a regex with Validator::make()
php
$validator = Validator::make(['agent' => $agent], ['agent' => ['required', 'regex:#^[a-zA-Z0-9 _.\-]+$#']]);
if ($validator->fails()) {
    return;
}
echo $agent;
// no issue — regex rule clears the taint
— Mechanism and API named with a working example v4.23.0
thinner coverage below
06 00
Insecure cookie flag detection Not checked yet new 50

Detects cookies created with the secure flag explicitly off as InsecureCookie when using response()->cookie().

— Names the check and API but no example given v4.23.0
07 00
Session value taint tracking Not checked yet new 50

Reads session values back from what the application stored under their key, so only tainted-at-source values remain flagged — getOldInput() and previousUrl() continue to be treated as user input.

— Mechanism described but no example or command v4.23.0
08 00
Minimum Psalm version raised to 7.13.1 Not checked yet breaking 50

Requires vimeo/psalm ^7.13.1; projects still on Psalm 7.12 must stay on v4.22.7.

— Exact version pin and migration note given v4.23.0
└──▷ BREAKING ON UPGRADE
  • !
    Requires vimeo/psalm ^7.13.1; projects on Psalm 7.12 must stay on v4.22.7.
Was this useful?

GitHub

Sources Blog / feed → Security advisory permissions are now generally available (2026-11-20) Blog Rank Why this rank top features scored 88/80/78, +15 breaking change, +6 runnable example — 103 How ranking works →

GitHub is a web-based platform for version control and collaborative software development using Git repositories.

GitHub's headline shipment is the GitHub Enterprise Server 3.23 release candidate, which bundles on-appliance AI secret detection, expiring OAuth tokens, wider malware advisory coverage, and a Dependabot cooldown, alongside GA security advisory permissions for custom roles, a Favorite repositories dashboard section, and multi-account support in GitHub Desktop.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 3 FEATURES9 FEATURES 4 OF 9 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
Expiring OAuth access tokens Not checked yet breaking 88

Apps can request expiring OAuth access tokensaccess tokensA credential that authorizes a client to use a service or API. It usually has limited permissions and a lifetime; anyone holding it may be able to use those permissions. with an 8-hour lifetime by passing scope=offline_access during authorization, or by requiring expiration in the app registration; refresh tokensrefresh tokensA credential used to obtain a new access token without signing in again. It is sent to the authorization service rather than used directly to call the protected API. can then obtain new access tokens.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Token expiration is now the default for newly created OAuth apps, and before requiring it on existing apps all clients must be updated to request, store, and use refresh tokens or sign-in will break.

— Names exact scope param and token lifetime, actionable default GitHub Enterprise Server 3.23 release candid…
02 00
Multiple accounts per host in GitHub Desktop Not checked yet new 80

GitHub Desktop now supports adding more than one account per host (GitHub.com, GitHub Enterprise Server, or GHE.com) with no fixed limit, via Settings/Options > Accounts. Each repository can be associated with a specific account so clone, publish, and Git operations automatically use the right credentials; users choose which signed-in account to use when a repo is accessible by multiple accounts, or Desktop infers it when only one applies, and the association can be changed later from repository settings or the repository list's context menu.

Add a second GitHub.com or Enterprise account so each repository automatically uses the right credentials without manual switching.
📍In GitHub Desktop 3.7.0, go to Settings/Options › Accounts and sign in with the additional account.
Re-assign an existing repository to a different signed-in account after adding multiple accounts.
📍Right-click the repository in the repository list and choose the new account from the context menu, or open the repository's settings page and change the account association.
— Full workflow and UI paths described, no CLI surface GitHub Desktop now supports multiple account…
03 00
Granular security advisory permissions for custom roles Not checked yet new 78

GitHub adds four security advisory permissions that can be assigned to custom repository roles under the Security section when creating or editing a role: View security advisories, Create, Edit, and Manage security advisories, 'Accept private vulnerability reports and close or reopen security advisories', and Publish security advisories and request CVEs. This lets orgs scope advisory access per-repository by assigning the role to users and teams only on the repositories they maintain, instead of granting full admin or org-wide security manager access.

Grant a triage team the ability to view and manage security advisories on a specific repo without making them admins or org-wide security managers
📍In your GitHub organization, go to Settings › Roles › New role, add the Create, Edit, and Manage security advisories permission under the Security section, then assign the role to your triage team on only the target repository under that repository's Settings › Collaborators and teams.
— Names all four permissions and exact UI path, no API endpoint Security advisory permissions are now genera…
04 00
Logical OR/AND filters in Projects views Not checked yet new 70

GitHub Projects views can now combine filters with logical OR, AND, and parentheses for grouping, for example (status:Done OR status:Todo) is:issue.

Narrow a GitHub Projects view to open issues that are either Done or Todo, using the new OR-grouping filter syntax.
$ (status:Done OR status:Todo) is:issue is:open
— Gives an exact runnable filter query GitHub Enterprise Server 3.23 release candid…
05 00
Favorite repositories section on dashboard Not checked yet new 65

Users can pin up to 25 repositories to the dashboard via the new Favorite repositories section, opened with Edit or Add repositories and saved with Save or discarded with Cancel. Favorites sync with pinned repositories in GitHub Mobile, with changes in either place appearing in the other after a refresh.

Pin the repositories you work in most often so they appear at the top of your GitHub dashboard.
📍On github.com, go to your Dashboard › Favorite repositories › select Edit or Add repositories › search for and select repositories › select Save.
— Exact UI path and 25-repo limit, no API/config key Pin your favorite repositories to your dashb…
06 02
On-appliance AI secret detection Not checked yet new 62

GHES can now detect unstructured secrets fully on-appliance, including in air-gapped environments, using AI secret detection that runs a fine-tuned model on the appliance CPUCPUThe main general-purpose processor in a computer. It runs application instructions and coordinates work with memory and other hardware. with no GPUGPUA graphics processing unit, a processor that performs many calculations in parallel. AI workloads use GPUs to speed up model training and inference., hosted LLMLLMA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code., or outbound connection required. The capability is in public preview.

— Explains mechanism and scope but no enablement steps GitHub Enterprise Server 3.23 release candid…
07 03
Malware advisory ingestion across seven ecosystems Not checked yet new 60

The GitHub Advisory Database now automatically ingests malware advisories from OpenSSF's malicious-packages repository, identifying known malicious packages in PyPI, Maven, RubyGems, NuGetNuGetMicrosoft's package manager for .NET, hosting a public registry of reusable libraries so tools can pull in dependencies without bundling or manually maintaining third-party code., Go, Rust, and Composer.

— Names all seven ecosystems and source feed, no action surface GitHub Enterprise Server 3.23 release candid…
thinner coverage below
08 04
Dependabot update cooldown period Not checked yet new 50

Dependabot now waits at least three days after a package version appears on its registry before opening a version-update pull request, giving maintainers time to detect compromised or buggy releases.

— Clear behaviour and duration, no config to adjust it GitHub Enterprise Server 3.23 release candid…
09 00
Dedicated node for GHES Redis primary Not checked yet improved 45

The Redis primary in a GHES cluster can now be moved to a dedicated node so it no longer shares resources with services such as MySQL.

— States the change, no migration commands given GitHub Enterprise Server 3.23 release candid…
└──▷ BREAKING ON UPGRADE
  • !
    Token expiration is the default for newly created OAuth apps; before requiring expiration on existing apps, all clients must be updated to request, store, and use refresh tokens or sign-in will break for clients that do not support them.
Was this useful?

syzkaller

Sources Commits → changes since 2026-11-17 Code

syzkaller is an unsupervised coverage-guided kernel fuzzer

syzkaller adds CSRFCSRFCross-site request forgery, an attack that makes a signed-in browser send an unwanted request to a site. The site may accept the request using the victim's existing session. protection to dashboard web UI forms and includes AI bisections in the bug lifetimes graph⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

  • syzkaller now adds CSRF protection to dashboard web UI forms, requiring a valid CSRFToken on state-changing POST requests and reading the token from csrfFormField.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Not checked yet
  • syzkaller now authenticates RPCRPCRemote Procedure Call, a protocol that lets a program execute a function on another machine as if it were local, giving cyber tools a structured way to communicate across processes or hosts without managing raw sockets. connections before unpacking large payloads in pkg/rpctype. Not checked yet
  • syzkaller's AI bisection flow now marks bisections that point to release commits as inconclusive rather than reporting them as results. Not checked yet
  • syzkaller's AI bisection backend now uses gemini-4-hydrogen as the core model, replacing gemini-3.6-flash. Not checked yet
  • syzkaller now falls back to the next LLMLLMA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code. model only on blocked requests, and drops thought signatures of other models on fallback. Not checked yet
  • syzkaller now adds CSRF protection to dashboard web UI forms, requiring a valid CSRFToken on state-changing POST requests and reading the token from csrfFormField.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Not checked yet
  • syzkaller now authenticates RPCRPCRemote Procedure Call, a protocol that lets a program execute a function on another machine as if it were local, giving cyber tools a structured way to communicate across processes or hosts without managing raw sockets. connections before unpacking large payloads in pkg/rpctype. Not checked yet

3 more in Everything

└──▷ BREAKING ON UPGRADE
  • !Dashboard admin and AI job handlers now require POST for state-changing UI actions; GET requests to those endpoints will no longer perform state changes.
Was this useful?
◆  Secrets Detection & Management

infisical

Sources Release notes → v0.167.14 Notes

Infisical is the open-source platform for securing secrets, certificates, access, and AI Agents.

Infisical adds bulk secret approval endpoints, AWS IAMAWS IAMAWS Identity and Access Management is Amazon's service for controlling who and what can call AWS APIs, by attaching policies that allow or deny actions on resources. Cyber tools use it to scope credentials to least privilege. gateway sessions beyond the 1-hour STSSTSAWS Security Token Service, an Amazon Web Services API that issues short-lived credentials for IAM roles, letting cyber tools authenticate to AWS without storing long-term secrets. cap, PKIPKIA framework of policies, roles, and cryptographic standards for issuing and managing digital certificates that bind public keys to identities, letting tools verify authenticity and encrypt communications without pre-shared secrets. certificate revocation approval, Slack alert channels, and grouped secret scan findings.

└──▷ GET THIS VERSION
$ git clone --branch v0.167.14 https://github.com/Infisical/infisical.git
# already have the repo? check out this version:
$ git checkout v0.167.14
  • Approve, reject, or merge secret approval requests in bulk with new batch review and merge API endpoints, including an approve-all and approve-without-merging option. Not checked yet
  • Broker AWS IAMIAMIdentity and Access Management — a framework of policies and services for controlling who can authenticate to systems and what resources they are authorized to use, letting cyber tools enforce least-privilege and audit access. CLI sessions through the gateway so they can outlive the 1-hour STS cap. Not checked yet
  • Require approval before revoking a PKI certificate. Not checked yet
  • Add auto-approval checks on the requester's user and groups for privileged access requests. Not checked yet
  • Secret scanningSecret scanningChecking files, code or other data for exposed credentials such as passwords, API keys and tokens. Findings help teams remove the secret and revoke or rotate it. now groups findings that share the same value. Not checked yet
+3 more show less
  • Configure Slack integration channels to receive alerts. Not checked yet
  • Org admins can now enforce multiple MFAMFAMulti-factor authentication, requiring proof from different categories, such as a password and a hardware key, to sign in. It reduces reliance on a password alone. methods on an organization. Not checked yet
  • Show the deployed CLI version on the gateway detail page. Not checked yet

8 details in Everything

Was this useful?

GitGuardian SaaS

Sources Blog / feed → Invalid secrets no longer block your developers and AI coding agents (2026-11-20) Blog Rank Why this rank top features scored 86/42, +6 runnable example — 70 How ranking works →

GitGuardian now filters confirmed-invalid secrets out of scan results so hooks and AI coding agents stop blocking on dead credentials, and Perimeter search gains repository URL lookup.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

GitGuardian SaaS is a secrets detection platform that scans Git repositories and developer workflows to identify and remediate exposed API keys, credentials, and sensitive data.

GitGuardian now filters confirmed-invalid secrets out of scan results so hooks and AI coding agents stop blocking on dead credentials, and Perimeter search gains repository URL lookup.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 2 FEATURES 0 OF 2 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Automatic filtering of confirmed-invalid secrets Not checked yet improved 86

GitGuardian's scan API now drops confirmed-invalid secrets from results for pre-commitpre-commitAn open-source framework for managing and running git hooks, maintained by the pre-commit project, that lets cyber tools enforce checks automatically on every commit before code reaches a repository., pre-push, ggshield secret scan ci, and direct calls to the /v1/scan and /v1/multiscan endpoints, so hooks and AI agents only block on valid or unverifiable secrets.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ The filter runs server-side, meaning no code change or ggshield upgrade is needed to benefit.

Verify that your CI pipeline now passes when only invalid secrets are detected — no code change or ggshield upgrade needed, since the filter runs server-side on the /v1/scan endpoint.
$ curl -s -X POST https://api.gitguardian.com/v1/scan \
  -H 'Authorization: Token <your-api-key>' \
  -H 'Content-Type: application/json' \
  -d '{"document": "<content-with-known-invalid-secret>", "filename": "config.yaml"}'
— Names exact endpoints, hooks and command, with runnable example Invalid secrets no longer block your develop…
thinner coverage below
02 00
Repository URL search in Perimeter Not checked yet new 42

Paste a full repository URL — including sub-page URLs for files or branches — into the Perimeter search or Source filter to locate the matching source.

— Clear UI starting point but no mechanism detail Invalid secrets no longer block your develop…
Was this useful?
◆  WAF, Bot & API Protection

Bunkerity BunkerWeb

Sources Release notes → v1.6.18 Notes Rank Why this rank no scored features — 0 How ranking works →

BunkerWeb is an open-source and cloud-native Web Application Firewall (WAF)

BunkerWeb adds Zstandard compression, per-service DNSBLDNSBLA DNS-based blocklist: a published list of IP addresses or domains associated with spam or abuse, queried via standard DNS lookups so cyber tools can check reputation in real time without downloading the full list. overrides, JSON ban output, URL-list refresh intervals, a new TLS ClientHelloTLS ClientHelloThe first message a TLS client sends when opening an encrypted connection, advertising supported cipher suites, TLS versions, and extensions. Cyber tools inspect or forge it to fingerprint clients or intercept traffic. plugin phase, and login throttling with configurable lockout settings.

└──▷ GET THIS VERSIONDockergit
Docker
$ docker pull ghcr.io/bunkerity/bunkerweb:1.6.18
git
$ git clone --branch v1.6.18 https://github.com/bunkerity/bunkerweb.git
# already have the repo? check out this version:
$ git checkout v1.6.18

published under the project's GitHub account · checked 21 Nov

  • Enable Zstandard response compression with USE_ZSTD, preferred over Brotli and gzip when the client accepts it; enabling it can also add Vary: Accept-Encoding to eligible uncompressed, gzip, and Brotli responses. Not checked yet
  • Set DNSBL_LIST per service so individual services can override the global DNS blocklist; services without an override keep the global list. Not checked yet
  • Output the current ban list as JSON for scripts and monitoring with bwcli bans --json. Not checked yet
  • Control how often URL lists are refreshed in the blacklist, greylist, and whitelist plugins with the new *_URLS_REFRESH_INTERVAL settings (default 1h). Not checked yet
  • Restrict antibot challenges to specific URIs with the new ANTIBOT_INCLUDE_URI setting; ANTIBOT_IGNORE_URI still takes precedence. Not checked yet
+11 more show less
  • Write ModSecurityModSecurityAn open-source web application firewall engine, maintained by the OWASP community, that inspects HTTP traffic against rule sets to block malicious requests before they reach an application. audit log entries as JSON objects with the new MODSECURITY_SEC_AUDIT_LOG_FORMAT setting (Native or JSON). Not checked yet
  • Run plugin code at the TLS ClientHello of every service through the new ssl_client_hello phase, ordered with PLUGINS_ORDER_SSL_CLIENT_HELLO. Not checked yet
  • Size the CrowdSec decision and metadata caches with CROWDSEC_CACHE_SIZE (default 100m) and CROWDSEC_METADATA_CACHE_SIZE (default 5m). Not checked yet
  • Limit BunkerNet reports to a fixed set of headers and blank matched values with BUNKERNET_FILTER_HEADERS, also applying retroactively to queued reports. Not checked yet
  • Override CLIENT_HEADER_TIMEOUT and CLIENT_BODY_TIMEOUT on a per-service basis. Not checked yet
  • Name a service by an IPv4 or IPv6IPv6The current version of the Internet Protocol, standardized by the IETF, using 128-bit addresses to replace the exhausted IPv4 space. Cyber tools target it because modern networks run dual-stack or IPv6-only. address with SERVER_NAME (e.g. SERVER_NAME=2001:db8::1); NGINX also matches the bracketed Host form. Not checked yet
  • Use utils.add_ban with opts.if_absent for insert-only bans that never replace an existing ban, locally or in Redis. Not checked yet
  • Export a support bundle with secrets redacted and IPs and domains anonymized using bwcli support. Not checked yet
  • New floating Docker image tags 1.6-edge, 1.6-rc, 1.6-beta, and 1.6-alpha track the latest release candidate or release for each channel. Not checked yet
  • The easy install script now routes downloads through HTTP_PROXY/HTTPS_PROXY and honours NO_PROXY. Not checked yet
  • New login throttling in the UI counts failed logins per client IP and per username, locking them for 15 minutes after five failures in 15 minutes, configurable with LOGIN_THROTTLE, LOGIN_MAX_ATTEMPTS, LOGIN_WINDOW_MINUTES, and LOGIN_LOCKOUT_MINUTES. Not checked yet

16 details in Everything

└──▷ BREAKING ON UPGRADE
  • !BunkerWeb ModSecurity rules move to IDs 9,800,000-9,809,999; custom exclusions referencing old IDs must be updated (mapping in the docs).
  • !Stream upstream TLS settings (REVERSE_PROXY_SSL_VERIFY=yes, protocols, ciphers, cached client certificate) now enable TLS to the backend instead of being ignored; clear them for a plaintext backend.
  • !Setting regexes are stricter: list values joined without a separator (e.g. 404500) and all-space values are refused; AUTH_BASIC_USER is now wider and accepts any value without :.
  • !The Lua API disk fallback now requires an explicit API_TOKEN line in variables.env; a file without one is refused.
Was this useful?

DataDome

Sources Blog / feed → iOS SDK for URLSession 1.0.0 (2026-11-20) Blog Rank Why this rank top features scored 75 — 75 How ranking works →

DataDome shipped a new iOS SDK that extends bot/fraud protection to apps using URLSession, with challenge handling and support for modern Swift concurrency patterns.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

DataDome protects web, mobile, and API applications by detecting and blocking bot and fraud traffic.

DataDome shipped a new iOS SDK that extends bot/fraud protection to apps using URLSession, with challenge handling and support for modern Swift concurrency patterns.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 1 FEATURE 1 OF 1 FEATURE most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
iOS SDK for URLSession Not checked yet new 75

A new ProtectedURLSession class intercepts and protects data, upload, download, and bytes URLSession task types on iOS.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Challenges returned by a redirect destination are retried at that destination, matching URLSession's redirected-request behaviour, and the SDK supports async/await, completion handlers, and Combine as calling conventions for all protected request types.

— Names exact class and task types but no install/setup steps given iOS SDK for URLSession 1.0.0 (2026-11-20)
Was this useful?
◆  Web Application Security

humble

Sources Release notes → 1.69 Notes Rank Why this rank top features scored 75/50/20, +10 product media, +6 runnable example — 64.33 How ranking works →

humble 1.69 adds safer non-interactive CI/CD behaviour and extends insecure cookie and header detection.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

A humble and fast security-oriented HTTP headers analyzer.

humble 1.69 adds safer non-interactive CI/CD behaviour and extends insecure cookie and header detection.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 3 FEATURES 0 OF 3 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Non-interactive mode via `-e yes` Not checked yet new 75

humble adds a yes value for the -e option, allowing it to run non-interactively in CI/CD pipelines where stdin is closed without crashing, e.g.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ python humble.py -u https://example.com -e yes.

Run humble non-interactively in a CI/CD pipeline where stdin is closed, using -e yes to avoid crashes.
$ python humble.py -u https://example.com -e yes
— Names exact flag and runnable command 1.69
thinner coverage below
02 00
Insecure `Partitioned` cookie detection Not checked yet new 50

humble now detects Partitioned cookies that are set without the Secure attribute and reports them as insecure.

— Names exact cookie attribute checked 1.69
03 00
Three new insecure header checks Not checked yet new 20

humble adds 3 new insecure header checks, though the specific headers are not named in the changelog.

— Count given but headers unnamed 1.69
Was this useful?
◆  Software Supply Chain Security

Socket

Sources Blog / feed → Expanded Java Coverage and More Accurate Reachability Results (2026-11-19) Blog Rank Why this rank top features scored 80/75/73, +15 breaking change, +6 runnable example — 97 How ranking works →

Socket fights vulnerabilities and provides visibility, defense-in-depth, and proactive supply chain protection for JavaScript, Python, and Go dependen...

Socket's biggest window changes are under the hood: expanded Java/Android reachability coverage, new webhook delivery-failure alerts, proxy support for reachability scans, and a reworked reachability-engine distribution mechanism and SBOMSBOMA software bill of materials, an inventory of the components and versions inside an application. It helps teams find affected software when a dependency has a vulnerability or licensing issue. export format that both carry breaking changes for existing integrations.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 10 FEATURES 3 OF 10 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
Reachability engine distribution and version change Not checked yet breaking 80

Full application reachability scans now require outbound access to api.socket.dev and storage.googleapis.com, since Socket downloads the reachability engine directly instead of through npm; SOCKET_CLI_COANA_LAUNCHER no longer takes effect and an npm mirror alone is no longer sufficient.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Older reachability engine versions released before this change cannot be used with --reach-version, and the engine itself is updated to Coana 15.15.0.

— Exact domains, env var and flag named but needs infra update Socket Basics 3.5.0: Reachability and Proxy …
02 00
Webhook failure notifications and auto-disable Not checked yet new 75

Socket now emails organization Owners and Admins when every outbound webhook delivery has failed for 12 hours, including the scheduled disable date, and automatically disables the webhook after four days of continuous failures with a follow-up email. The webhooks table also shows Unhealthy and Deliveries failing status badges so teams can track and re-enable affected endpoints.

Re-enable a webhook after fixing its endpoint so previous failures no longer count toward automatic disabling.
📍In the Socket dashboard, go to Settings › Webhooks, find the entry marked Deliveries failing, fix the endpoint URL or configuration, then set the webhook status back to Active.
— Clear thresholds and a UI path to re-enable webhooks Get Notified When Webhook Deliveries Fail (2…
03 00
Argument handling improvements for reachability scans Not checked yet improved 73

Reachability scans accept repeated --exclude-paths flags to supply multiple path exclusions, and socket_additional_params now preserves commas and quoted spaces within values so options such as --reach-ecosystems=npm,pypi are passed through correctly.

— Exact flags and parameter names a reader can apply directly Socket Basics 3.5.0: Reachability and Proxy …
04 02
SPDX export relationship and supplier fields Not checked yet breaking 73

Socket's SPDXSPDXAn open standard for describing software components, licenses and their relationships. Its documents let different tools exchange a software bill of materials without inventing a new format. exports now use DEPENDS_ON relationships (from project to dependencies) instead of DEPENDENCY_OF, and include supplier information or NOASSERTION when the publisher is unknown. Integrations reading DEPENDENCY_OF relationships must be updated to read DEPENDS_ON instead.

— Exact field names given though no new endpoint Improved SPDX and CycloneDX SBOM Exports (20…
05 00
Java and Android reachability coverage expansion Not checked yet new 65

Socket's reachability engine now analyzes Android libraries packaged as AARs and traces Spring Boot auto-configuration, Spring XML bean definitions, service loading, JPA and servlet descriptors, and calls from Java into Scala objects to compute full application reachability.

— Names several specific mechanisms but no flag or command to invoke it Expanded Java Coverage and More Accurate Rea…
06 03
Reachability timeout flag syntax change Not checked yet breaking 65

The reachability scan timeout flag now uses the equals-sign form --reach-analysis-timeout=90s instead of the comma form --reach-analysis-timeout,90s, which is no longer correct.

Run a reachability scan covering both npm and PyPI ecosystems behind a corporate proxy, with a 90-second timeout and specific paths excluded.
$ HTTPS_PROXY=https://proxy.corp.example:8080 socket scan --reach-ecosystems=npm,pypi --reach-analysis-timeout=90s --exclude-paths=test/ --exclude-paths=vendor/
— Exact before and after flag syntax given Socket Basics 3.5.0: Reachability and Proxy …
07 00
Platform-specific Ruby gem handling in SBOM exports Not checked yet breaking 65

Platform-specific Ruby gems (such as nokogiri) locked in Gemfile.lock now appear once per platform, each with the registry version and a platform qualifier, so downstream tools can match them correctly. The first diff scan after this update will show affected platform-specific gems as removed and re-added once.

— Names exact file and example package but no UI path given Improved SPDX and CycloneDX SBOM Exports (20…
08 00
Proxy support for reachability scans Not checked yet new 60

Reachability scans now honour the HTTPS_PROXY and NO_PROXY environment variables, enabling scans to run behind a corporate proxy.

Run a reachability scan covering both npm and PyPI ecosystems behind a corporate proxy, with a 90-second timeout and specific paths excluded.
$ HTTPS_PROXY=https://proxy.corp.example:8080 socket scan --reach-ecosystems=npm,pypi --reach-analysis-timeout=90s --exclude-paths=test/ --exclude-paths=vendor/
— Named env vars with a runnable example command Socket Basics 3.5.0: Reachability and Proxy …
thinner coverage below
09 00
Reachability accuracy and performance improvements Not checked yet improved 50

Vulnerabilities in libraries Socket cannot analyze now show Unknown instead of an incorrect Unreachable, falling back to precomputed results when a vulnerability pattern does not match the package. Projects using multiple versions of a crate now get call stacks specific to each version rather than a shared stack, and reachability analysis performance and precision improved for JavaScript, TypeScript and Python projects (with particular gains for PyTorch) as well as for Go.

— Describes engine behaviour changes with no actionable command Expanded Java Coverage and More Accurate Rea…
10 00
CycloneDX export includes direct dependencies Not checked yet improved 35

Socket's CycloneDXCycloneDXAn open standard for describing software components and their dependencies, vulnerabilities and other supply-chain information. Tools use it to exchange bills of materials and related evidence. exports now include the project's direct dependencies.

— Thin one-line description with no further detail Improved SPDX and CycloneDX SBOM Exports (20…
└──▷ BREAKING ON UPGRADE
  • !
    SOCKET_CLI_COANA_LAUNCHER no longer takes effect; Socket downloads the reachability engine directly and an npm mirror alone is no longer sufficient.
  • !
    Older reachability engine versions released before this change cannot be used with --reach-version.
  • !
    Full application reachability scans now require outbound access to api.socket.dev and storage.googleapis.com; existing setups that relied on an npm mirror must be updated.
  • !
    The comma form --reach-analysis-timeout,90s is no longer correct; use --reach-analysis-timeout=90s instead.
  • !
    If your integration reads DEPENDENCY_OF relationships from Socket's SPDX exports, update it to read DEPENDS_ON instead.
  • !
    The first diff scan after this update will show affected platform-specific Ruby gems as removed and re-added once.
Was this useful?
Reverse Engineering, Malware & Forensics
◆  Malware Analysis

YARA-X

Sources Release notes → v1.24.0 Notes Rank Why this rank top features scored 78/68/57, +6 runnable example — 73.67 How ranking works →

YARA-X v1.24.0 expands the DEXDEXA decentralized exchange protocol running on a blockchain that lets users trade tokens directly from their wallets via smart contracts, removing the need for a centralized custodian. module with new containment-check functions, adds Windows ARM64ARM64A 64-bit CPU architecture from ARM Holdings, used in mobile chips, Apple Silicon, and cloud servers; cyber tools target it to run natively on those platforms instead of relying on emulation. release builds and man pages, and improves scanner memory efficiency and rule complexity limits.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

YARA-X is a rewrite of YARA in Rust.

YARA-X v1.24.0 expands the DEXDEXA decentralized exchange protocol running on a blockchain that lets users trade tokens directly from their wallets via smart contracts, removing the need for a centralized custodian. module with new containment-check functions, adds Windows ARM64ARM64A 64-bit CPU architecture from ARM Holdings, used in mobile chips, Apple Silicon, and cloud servers; cyber tools target it to run natively on those platforms instead of relying on emulation. release builds and man pages, and improves scanner memory efficiency and rule complexity limits.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 7 FEATURES 0 OF 7 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
DEX module string, method and class checks Not checked yet new 78

The dex module gains has_string, has_method, and has_class functions to test for the presence of strings, methods, or classes in DEX files; the previous names contains_string, contains_method, and contains_class are kept as aliases for YARA compatibility.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Query a DEX file for a specific class using the new canonical function name in a YARA-X rule.
yara
rule dex_example {
  condition:
    dex.has_class("Lcom/example/MaliciousClass;")
}
— Names exact functions and aliases with a runnable rule example. v1.24.0
02 00
Higher WASM instruction limit for rule conditions Not checked yet improved 68

The maximum WASMWASMA binary instruction format for a stack-based virtual machine, maintained by the W3C, that lets code compiled from any language run at near-native speed inside a sandboxed environment across browsers and servers. function instruction limit (MAX_FUNC_INSTRS) was increased from 100,000 to 300,000, allowing larger and more complex rule conditions to compile.

— Gives exact before/after limit values. v1.24.0
thinner coverage below
03 00
Alias argument for module_export macro Not checked yet new 57

The #[module_export] macro now supports an optional alias argument, letting module authors export functions under alternate names.

— Names the macro and argument but no usage example. v1.24.0
04 00
Shared regex compilation across scanners Not checked yet improved 55

Compiled regular expressions and regexp sets are now shared across Scanner instances within Rules, reducing memory use when running many scanners against the same rule set.

— Explains mechanism and benefit but no concrete numbers. v1.24.0
05 00
Generated man pages for yr CLI Not checked yet new 50

Man pages for yr and its subcommands are now generated and included under cli/man/.

— Names exact location but describes only the addition itself. v1.24.0
06 00
Windows ARM64 release builds Not checked yet new 45

YARA-X now ships Windows ARM64 release builds for the CLI, the C API, and Python wheels.

— Names the exact artifacts but no further mechanism. v1.24.0
07 00
Full keyword highlighting in Playground editor Not checked yet improved 30

The Playground editor now highlights all YARA keywords and operators.

— Minor UI change described only briefly. v1.24.0
Was this useful?

CAPEv2

Sources Commits → changes since 2026-11-19 Code

CAPEv2 is an open-source malware sandbox that automatically executes and analyzes suspicious files to extract configurations, payloads, and behavioral indicators.

CAPEv2 adds a Granian ASGIASGIA Python async web server interface specification that defines how servers and applications exchange HTTP and WebSocket messages; cyber tools use it to serve high-concurrency APIs without blocking on I/O. server, migrates the host runtime from Poetry to uv with Python 3.14, and adds a migrate-uv installer command⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ TRY IT
Migrate an existing CAPEv2 host from Poetry to uv and Python 3.14 without reinstalling the sandbox.
$ sudo bash installer/cape2.sh migrate-uv | tee migrate-uv.log
  • Migrate an existing Poetry-based host to uv with Python 3.14 in-place by running sudo bash installer/cape2.sh migrate-uv (documented in docs/book/src/installation/uv_migration.rst), installing a standalone CPython 3.14 interpreter in /usr/local/share/uv/python without touching the system /usr/bin/python3.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Not checked yet
  • CAPEv2's web and ASGI stack now uses Granian, dropping the previous daphne and uWSGI servers. Not checked yet
  • The --use-uv installer flag is now a no-op; uv with Python 3.14 is the default for all new installs. Not checked yet
  • Migrate an existing Poetry-based host to uv with Python 3.14 in-place by running sudo bash installer/cape2.sh migrate-uv (documented in docs/book/src/installation/uv_migration.rst), installing a standalone CPython 3.14 interpreter in /usr/local/share/uv/python without touching the system /usr/bin/python3.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Not checked yet
  • CAPEv2's web and ASGI stack now uses Granian, dropping the previous daphne and uWSGI servers. Not checked yet

1 more in Everything

└──▷ BREAKING ON UPGRADE
  • !The --use-uv installer flag is now a deprecated no-op; uv with Python 3.14 is the default. Existing Poetry-based hosts must run sudo bash installer/cape2.sh migrate-uv to migrate.
  • !The web/ASGI stack has dropped daphne and uWSGI in favour of Granian; any systemd or proxy configuration referencing daphne or uWSGI must be updated.
Was this useful?
Cloud & Container Security
◆  Container & Image Security

Chainguard

Sources Changelog → changelog · Nov 21, 2026 Changelog Rank Why this rank top features scored 75/65/60, +10 product media — 76.67 How ranking works →

Chainguard opened self-serve access with new Starter and Team plans, launched the Rust-based Project Theseus initiative to rebuild core open-source components starting with a new memory allocator, disclosed a second wave of silent Java vulnerabilities via Athena, and added 25 new container images.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Chainguard provides trusted open source artifacts for every layer of your modern software stack—containers, language libraries, and VM images.

Chainguard opened self-serve access with new Starter and Team plans, launched the Rust-based Project Theseus initiative to rebuild core open-source components starting with a new memory allocator, disclosed a second wave of silent Java vulnerabilities via Athena, and added 25 new container images.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 4 FEATURES 2 OF 4 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Athena discloses 12 more silent Java vulnerabilities Not checked yet new 75

Athena's second wave discloses 12 more silent vulnerabilities in Java projects, including two critical authentication bypasses across nine libraries.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Remediated versions are available from Chainguard Repository using the -0cgr.n version qualifier, adoptable via a lockfile change and rebuild.

— Names exact version qualifier and remediation path changelog-20261121-fc0559fe
02 01
Project Theseus and the challoc allocator Not checked yet new 65

Project Theseus is a new Chainguard initiative to rebuild core open-source components in Rust to eliminate classes of vulnerabilities rather than patching them repeatedly. The first component is challoc, a memory allocator replacing glibc malloc, jemalloc, tcmalloc, and mimalloc.

— Clear mechanism and named component, but no adoption steps changelog-20261121-fc0559fe
03 02
Self-serve Starter and Team plans Not checked yet new 60

Chainguard now offers self-serve sign-up with two new plans: free Chainguard Starter (no credit card required) giving access to Chainguard Containers, Chainguard Libraries, and other products within usage limits, and paid Chainguard Team which raises most Starter limits and adds ticket support.

— Names plans and scope but no exact limits or pricing changelog-20261121-fc0559fe
thinner coverage below
04 00
25 new container images added Not checked yet new 50

Chainguard added 25 new container images this week, including aws-route53-controller, grafana-pyroscope-fips, dify-web, kata-monitor, kgateway-envoy-wrapper, kgateway-sds, neuvector-registry-adapter, oc-mirror, and gosu-fips, among others.

— Lists concrete image names but no usage detail changelog-20261121-fc0559fe
Was this useful?
◆  SaaS Security Posture

Nudge Security

Sources Release page → Get findings for AI agents discovered in the browser Notes

Nudge Security is a SaaS security platform that discovers shadow IT assets, manages identity governance, and hardens security posture.

Nudge Security now surfaces AI agent findings from the browser extension in the same Findings view as connected-app agent findings.

  • New detectionNudge Security now generates findings for AI agents discovered through the browser extension, surfacing them in the Findings view alongside findings for agents discovered through connected apps. Not checked yet
  • Enable AI agent discovery in the browser extension settings to ensure browser-discovered agents are included in findings. Not checked yet

2 details in Everything

Was this useful?
◆  Cloud Security Posture

Sysdig Secure

Sources Blog / feed → Secure SaaS - November 20, 2026 Blog Rank Why this rank top features scored 62, +10 product media, +6 runnable example — 78 How ranking works →

Sysdig Secure adds a Stage filter to its vulnerability views, letting teams see registry, pipeline, and runtime scan results together or separately.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Sysdig Secure is a cloud-native security platform that provides runtime threat detection, vulnerability management, and compliance monitoring for containers, Kubernetes, and cloud environments.

Sysdig Secure adds a Stage filter to its vulnerability views, letting teams see registry, pipeline, and runtime scan results together or separately.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 1 FEATURE 0 OF 1 FEATURE most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Stage filter for vulnerability findings Not checked yet new 62

A new Stage filter on the Vulnerability Findings page and the Vulnerability Overview dashboard lets users filter by Pipeline, Registry, Runtime, or any combination, and both views now include registry and pipeline scan results alongside runtime findings.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Review all pre-runtime (registry and pipeline) vulnerability findings alongside runtime findings in one view, filtered to only pipeline stage results.
📍In Sysdig Secure, go to Vulnerabilities › Findings, then open the Stage filter and select Pipeline (or Registry, Runtime, or any combination) to scope the results.
— Names exact filter values and two UI locations, but no mechanism detail Secure SaaS - November 20, 2026
Was this useful?
Detection & Response
◆  Endpoint Security (EDR/XDR)

Huntress

Sources Release page → Control how Huntress notifications are prioritized in HaloPSA Notes Rank Why this rank top features scored 72/35, +6 runnable example — 59.5 How ranking works →

Huntress shipped two unrelated improvements this window: configurable HaloPSA ticket priority mapping and a simpler sign-in flow for Huntress MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface..⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Give feedback to the Huntress team so we can make more informed product decisions. Powered by Canny.

Huntress shipped two unrelated improvements this window: configurable HaloPSA ticket priority mapping and a simpler sign-in flow for Huntress MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface..⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 2 FEATURES 0 OF 2 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Notification-severity-to-ticket-priority mapping for HaloPSA Not checked yet new 72

Admins can map each Huntress notification severity (e.g.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Critical, High, Medium, Low) to a HaloPSA ticket priority, giving control over how Huntress-generated tickets are prioritized in HaloPSA. Configured in the Huntress portal under Settings > Integrations > HaloPSA, under Ticket Priority Mapping.

Set ticket priority in HaloPSA to match your team's response SLAs for each Huntress alert severity level.
📍In the Huntress portal, go to Settings › Integrations › HaloPSA, then under Ticket Priority Mapping assign a HaloPSA ticket priority to each Huntress notification severity (e.g. Critical, High, Medium, Low).
— Names exact UI path and mapping mechanism Control how Huntress notifications are prior…
thinner coverage below
02 00
Portal credential sign-in for Huntress MCP Not checked yet improved 35

Users can sign in to Huntress MCP directly with their Huntress portal credentials, bypassing the need for an administrator to first create an API key.

— Describes benefit but no concrete steps given Sign in to Huntress MCP with your portal cre…
Was this useful?
◆  SIEM

Elastic Security Labs

Sources Blog / feed → Mapping the AI threat landscape: MITRE ATLAS comes to Elastic Security (2026-11-20) Blog

Elastic Security Labs is a powered by Elasticsearch. Ready to deploy your data?

Elastic Security 9.6 adds a new ATLAS coverage overview table, ATLAS support in the rule form and AI rule creation workflow, Agent Builder ATLAS awareness, and a new managed MITREMITREA US federally funded research and development center that maintains public frameworks—ATT&CK, CVE, D3FEND—that cyber tools use to classify threats, vulnerabilities, and defensive techniques in a common language. data service in Kibana.

  • A new ATLAS coverage overview table shows which ATLAS tactics and techniques your rules already address, scored in its own view separate from ATT&CKATT&CKA publicly maintained knowledge base by MITRE cataloguing adversary tactics, techniques, and procedures observed in real attacks, giving cyber tools a common taxonomy for classifying and correlating threat behavior. numbers. Not checked yet
  • Updated detectionThe detection rule creation and edit form now supports adding both ATT&CK and ATLAS mappings to custom rules or existing prebuilt rules. Not checked yet
  • The AI-assisted rule creation workflow now supports adding both ATT&CK and ATLAS mappings to rules. Not checked yet
  • Agent Builder skills now understand ATLAS, letting you ask an agent which AI threats your installed rules cover or have it recommend prebuilt rules for specific ATLAS tactics and techniques. Not checked yet
  • A new managed MITRE data service in Kibana ingests MITRE ATLASMITRE ATLASA MITRE-maintained knowledge base of adversarial tactics and techniques targeting AI and machine learning systems, modelled on ATT&CK, giving cyber tools a structured taxonomy for AI-specific threat modelling. and ATT&CK entities as Saved Objects, enabling querying and filtering via a data client and API, and decoupling framework updates from Kibana release cycles. Not checked yet

5 details in Everything

Was this useful?

Databricks

Sources Blog / feed → MCP tools in Genie Agents are generally available (2026-11-20) Blog Rank Why this rank top features scored 90/70/70, +10 product media, +6 runnable example — 92.67 How ranking works →

Databricks shipped GA warehouse-level statement timeouts with full API/Terraform/SDK support, GA Delta SharingDelta SharingAn open protocol, maintained by the Linux Foundation, for sharing live datasets across organizations without copying them, letting cyber tools query remote data in place over standard REST and cloud storage. across regulatory domains, GA MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools for Genie Agents, a new Lakewatch CLI command group, GitHub app pull-request automation, and GA LTAP Direct Writes for synced tables.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Databricks is a unified data analytics platform that enables collaborative data engineering, machine learning, and analytics using Apache Spark in the cloud.

Databricks shipped GA warehouse-level statement timeouts with full API/Terraform/SDK support, GA Delta SharingDelta SharingAn open protocol, maintained by the Linux Foundation, for sharing live datasets across organizations without copying them, letting cyber tools query remote data in place over standard REST and cloud storage. across regulatory domains, GA MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools for Genie Agents, a new Lakewatch CLI command group, GitHub app pull-request automation, and GA LTAP Direct Writes for synced tables.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 6 FEATURES 0 OF 6 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Warehouse-level statement timeouts Not checked yet new 90

Set per-warehouse query time limits using the statement_timeout field across the SQL warehouses API, Databricks SDKs, Terraform, or Declarative Automation Bundles to automatically cancel runaway statements.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ This is now generally available and can be set on a Terraform databricks_sql_warehouse resource or applied directly via PATCH /api/2.0/sql/warehouses/<warehouse-id>.

Cap long-running queries on a shared warehouse so runaway jobs cannot monopolize compute — set statement_timeout in your Terraform warehouse resource.
hcl
resource "databricks_sql_warehouse" "shared" {
  name = "shared-warehouse"
  statement_timeout = 3600
}
Apply a statement timeout directly via the SQL warehouses API to an existing warehouse without changing IaCIaCInfrastructure as Code: the practice of defining and provisioning compute infrastructure through machine-readable config files rather than manual processes, giving cyber tools repeatable, version-controlled environment setup..
$ curl -X PATCH https://<workspace-host>/api/2.0/sql/warehouses/<warehouse-id> \
  -H 'Authorization: Bearer <token>' \
  -H 'Content-Type: application/json' \
  -d '{"statement_timeout": 3600}'
— Names exact field, API endpoint, and runnable Terraform/curl examples Warehouse-level statement timeouts are now g…
02 00
Lakewatch management via Databricks CLI Not checked yet new 70

Manage Lakewatch datasources, parsers, detection rules, and per-catalog configuration from the command line using the new lakewatch command group in the Databricks CLI, now in beta. Requires CLI version 1.25.0 or above, with parser commands requiring version 1.27.0 or above.

— Names command group and version gates but no sample invocation Manage Lakewatch with the Databricks CLI (Be…
03 00
LTAP Direct Writes for synced tables Not checked yet new 70

Enable LTAP Direct Writes when creating a synced table so loads write directly to Lakebase storage instead of through your compute endpoint, making large initial loads and full refreshes finish faster without competing with application queries. Now generally available, selectable under Sync method when creating a synced table in the Databricks UI.

When creating a synced table with a large initial data load, select LTAP Direct Writes so the ingestion goes straight to Lakebase storage and does not slow down running application queries.
📍In the Databricks UI, go to Lakebase › Synced Tables › Create Synced Table, then under Sync method select LTAP Direct Writes before saving.
— Explains mechanism and gives UI navigation steps LTAP Direct Writes for synced tables is now …
04 00
Delta Sharing across regulatory domains Not checked yet new 60

Databricks-to-Databricks Delta Sharing across regulatory domains is now generally available, letting commercial workspaces share data with AWS GovCloud, AWS GovCloud DoD, and Azure China environments.

— Names target domains but no setup steps or command Databricks-to-Databricks Delta Sharing acros…
thinner coverage below
05 00
MCP tools in Genie Agents Not checked yet new 55

Genie Agents can now attach external MCP servers registered in Unity Gateway, in addition to Databricks-provided connectors, and Agent mode APIs now support MCP tools. This capability is now generally available.

— Names Unity Gateway and Agent mode APIs but no config example MCP tools in Genie Agents are generally avai…
06 00
GitHub app pull request automation Not checked yet new 55

Genie Code and Genie ZeroOps can now create and manage pull requests on your behalf once the Databricks GitHub app is granted Read and write access to pull requests; existing branch protection rules and required reviews still apply.

— Names the permission and products but no setup command The Databricks GitHub app will request permi…
Was this useful?
◆  Detection Engineering

Tenzir

Sources Blog / feed → Tenzir v7.0.0-rc.7: A new foundation for security data (2026-11-20) Blog Pre-release

Tenzir is a data pipeline engine for security teams that collects, transforms, enriches, and routes telemetry.

Tenzir v7.0.0-rc.7 adds an MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server for AI agents, a built-in assistant, tenzir help, from_splunk for SPL2SPL2Splunk's second-generation query and processing language, replacing SPL, offering a pipeline syntax for searching, transforming, and routing data across Splunk's platform. searches, and a rebuilt deployment model with versioned packages and live rollout.

  • Run SPL2 searches against Splunk data with the new from_splunk operator. Not checked yet
  • Get inline help at any time with the new tenzir help command. Not checked yet
  • Tenzir now includes an MCP server so AI agents can interact with the pipeline. Not checked yet
  • Tenzir now includes a built-in assistant for guided pipeline authoring. Not checked yet
  • Tenzir's platform and deployment model is rebuilt with configuration management, versioned packages, and live rollout. Not checked yet
+6 more show less
  • Homebrew installation now puts tenzir on the PATH automatically. Not checked yet
  • Native Hydrolix semantics are now applied to operators after from_hydrolix, enabling tighter pushdown. Not checked yet
  • Native Microsoft GraphMicrosoft GraphMicrosoft's unified REST API for accessing data and services across Microsoft 365, Azure AD, and related cloud products, giving cyber tools a single authenticated endpoint to query users, emails, files, and sign-in logs. semantics are now applied to operators after from_microsoft_graph. Not checked yet
  • Native CloudWatch Logs semantics are now applied to operators after from_amazon_cloudwatch. Not checked yet
  • Content-skipping filters and limits are now supported after from_microsoft_365_activity. Not checked yet
  • Operators after from_velociraptor now stay in the pipeline by design rather than exiting early. Not checked yet

11 details in Everything

└──▷ BREAKING ON UPGRADE
  • !Platform and deployments have been reimagined — existing deployment configurations may not carry over without migration.
Was this useful?
Exposure & Vulnerability Management
◆  Endpoint Management (UEM/RMM)

Devicie

Sources Press release → announced · Nov 21, 2026 Press

Devicie is a keep pace with application updates and configuration changes. Devicie automates recurring management natively through Microsoft Intune.

Devicie adds multi-tenant MSPMSPA managed service provider, a company that operates IT services for customers. It may manage systems, networks, backups or other ongoing technology needs. portal support, letting MSPsMSPsA managed service provider, a company that operates IT services for customers. It may manage systems, networks, backups or other ongoing technology needs. view browser security events across all customer tenants in one place with per-tenant reporting keys.

Was this useful?
◆  Security Validation (BAS & Automated Pentest)

Horizon3 NodeZero

Sources Release page → updated · Nov 21, 2026 Notes Rank Why this rank top features scored 45/45/40, +10 product media — 53.33 How ranking works →

NodeZero's latest update focuses on expanding testing coverage and multi-tenant/enterprise administration, adding REST APIREST APIA web API design style where clients interact with resources over HTTP using standard methods like GET, POST, and DELETE, giving cyber tools a predictable, language-agnostic way to integrate with external services.-only webapp testing, an Audit Log dashboard, Kubernetes pentest configuration options, and a Tripwires webhook connector for SIEMSIEMSecurity information and event management, collecting logs from many systems and correlating events to find suspicious activity. Analysts use it to investigate alerts and incidents./SOARSOARSecurity orchestration, automation and response, coordinating incident workflows across tools. Playbooks can gather evidence, assign work or carry out approved response actions. integration.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Horizon3 NodeZero is a horizon3.ai enables organizations to continuously assess the security posture of their enterprise across many attack surfaces.

NodeZero's latest update focuses on expanding testing coverage and multi-tenant/enterprise administration, adding REST APIREST APIA web API design style where clients interact with resources over HTTP using standard methods like GET, POST, and DELETE, giving cyber tools a predictable, language-agnostic way to integrate with external services.-only webapp testing, an Audit Log dashboard, Kubernetes pentest configuration options, and a Tripwires webhook connector for SIEMSIEMSecurity information and event management, collecting logs from many systems and correlating events to find suspicious activity. Analysts use it to investigate alerts and incidents./SOARSOARSecurity orchestration, automation and response, coordinating incident workflows across tools. Playbooks can gather evidence, assign work or carry out approved response actions. integration.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 7 FEATURES 1 OF 7 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
Kubernetes pentest configuration options Not checked yet new 45

Kubernetes pentests can now be configured with custom service accounts, by importing rules from YAML files, and by setting up a proxy.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Names three config options but no exact flags or keys snapshot-20261121
02 00
Client Management for parent accounts Not checked yet new 45

Parent accounts in multi-tenant environments now have a top-level view of clients' operational health, license allocations, and API keys (Client Management).

— Describes scope of new view but no navigation detail snapshot-20261121
03 00
Tripwires webhook connector Not checked yet new 40

Tripwires alerts can now be sent to SIEM, SOAR, or automation platforms through a new webhook connector.

— Describes destination types but no endpoint or setup details snapshot-20261121
04 00
Audit Log dashboard for Org Admins Not checked yet new 40

Org Admins now have a dashboard to review, investigate, and export records of account activity (Audit Log).

— Names dashboard and export capability, no exact UI path snapshot-20261121
05 00
REST API-only webapp testing Not checked yet new 35

The WebApp pentest engine can now test REST APIsREST APIsA convention for building HTTP-based interfaces where resources are addressed by URL and manipulated with standard verbs like GET, POST, and DELETE, giving cyber tools a predictable way to integrate with external services. that don't expose a web UI.

— Clear capability but no mechanism or setup steps given snapshot-20261121
06 00
Parallel WebApp pentest execution with expanded evidence Not checked yet improved 35

WebApp pentest flows now run in parallel, returning faster results, alongside expanded adaptive agents and verifiable evidence included in reports.

— Explains behaviour change but no metrics or specifics snapshot-20261121
07 00
Updated BloodHound integration Not checked yet improved 10

The BloodHound integration is now updated to a newer version.

— No version number or change details given snapshot-20261121
└──▷ ALSO FROM THESE RELEASES
Was this useful?
◆  Vulnerability & Exposure Management

oxo

Sources Release notes → v2.25.4 Notes Rank Why this rank top features scored 45/45 — 45 How ranking works →

OXO is a security scanning orchestrator for the modern age.

oxo extended two message schemas this window, adding certificate history data to fingerprinting and a capture timestamp to UI call tracking.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 2 FEATURES 0 OF 2 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Certificate data in fingerprint history message Not checked yet improved 45

Adds x509x509A standard defining the format of public-key certificates, maintained by the ITU-T, used by cyber tools to authenticate identities and establish encrypted connections over TLS and similar protocols. certificate data to the v3.fingerprint.certificate.history message for richer certificate history fingerprinting.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Names exact message field but no mechanism detail. v2.25.4
02 00
Timestamp field on UI call capture message Not checked yet improved 45

Adds a captured_at field to the v3.capture.ui_call message to record when a UI call was captured.

— Names exact message field but no mechanism detail. v2.25.3
Was this useful?

Qualys VMDR

Sources Blog / feed → changelog · Nov 20, 2026 Blog Rank Why this rank top features scored 88, +6 runnable example — 94 How ranking works →

Qualys VMDR is a cloud-based platform that continuously discovers, assesses, prioritizes, and remediates vulnerabilities across an organization's IT assets.

Qualys VMDR added a TruRiskTruRiskA proprietary risk-scoring methodology from Qualys that translates vulnerability and asset data into a numeric business-risk score, giving cyber tools a single prioritization signal instead of raw CVSS values. Eliminate mitigation for Java4Shell that applies a JVMJVMThe Java Virtual Machine, a runtime maintained by Oracle (and implemented by others) that executes Java bytecode on any OS, giving cyber tools a portable, garbage-collected environment with a large ecosystem of libraries. flag to eligible hosts, reports process-level coverage, and supports one-click rollback.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 1 FEATURE 1 OF 1 FEATURE most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
Java4Shell mitigation via TruRisk Eliminate Not checked yet new 88

TruRisk Eliminate applies the -XX:-UseBASE64Intrinsics flag to JAVA_TOOL_OPTIONS on affected Linux and Windows hosts (linked to QIDs such as 389494), automatically checking each host for an AVX-512 VBMI CPUCPUThe main general-purpose processor in a computer. It runs application instructions and coordinates work with memory and other hardware. and an affected Java runtime before making any change.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ After mitigation, the console reports every running Java process with its effective JVM option setting in HotSpot resolution order, flagging unknown processes to surface coverage gaps, and the mitigation can be rolled back in one click, restoring the previous JAVA_TOOL_OPTIONS value.

When you need a bridge fix for Java4Shell before patched JDK builds are available, apply the JVM flag across all eligible hosts without touching containers.
📍In the Qualys console, go to TruRisk Eliminate, select the Java4Shell mitigation (linked to QIDs such as 389494), choose your target Linux or Windows hosts, and click Apply. The Cloud Agent checks each host for an AVX-512 VBMI CPU and an affected Java runtime, then sets -XX:-UseBASE64Intrinsics in JAVA_TOOL_OPTIONS. Plan JVM restarts with application owners; Windows services require a host reboot. For containers, set the flag in the image or pod spec manually.
— Names exact flag, config var, QIDs, and console workflow with rollback. 2026-11-20-a6009168
Was this useful?
Identity, Access & Data Protection
◆  Identity & Access Management

openfga

Sources Release notes → v1.23.0 Notes Rank Why this rank top features scored 83/73/65, +15 breaking change — 88.67 How ranking works →

OpenFGA v1.23.0 introduces a typed query ASTASTAn abstract syntax tree, a structured representation of parsed code. Tools inspect or transform this tree to understand expressions and statements rather than just matching text. and Querier interface for structured backend queries (with a SQLite implementation), optimizes PostgreSQL's latest-model lookup, and upgrades grpc-gatewaygrpc-gatewayA Go library that reads protobuf service definitions and generates a reverse proxy, translating RESTful HTTP/JSON requests into gRPC calls so tools can expose both APIs from one backend. with header and query-parameter validation changes.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Openfga is a high-performance and flexible authorization/permission engine built for developers and inspired by Google Zanzibar.

OpenFGA v1.23.0 introduces a typed query ASTASTAn abstract syntax tree, a structured representation of parsed code. Tools inspect or transform this tree to understand expressions and statements rather than just matching text. and Querier interface for structured backend queries (with a SQLite implementation), optimizes PostgreSQL's latest-model lookup, and upgrades grpc-gatewaygrpc-gatewayA Go library that reads protobuf service definitions and generates a reverse proxy, translating RESTful HTTP/JSON requests into gRPC calls so tools can expose both APIs from one backend. with header and query-parameter validation changes.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 3 FEATURES 2 OF 3 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
Typed query AST and Querier interface for storage backends Not checked yet breaking 83

Adds a typed query AST and new Querier interface in pkg/storage/adapter, letting custom RelationshipTupleReader backends run queries as structured statements instead of hand-built SQL; OpenFGA now implements Querier for SQLite, enabling future SQL optimizations for SQLite-backed deployments.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Custom RelationshipTupleReader implementations (library use only) must now implement a Querier method on the interface — returning nil is safe but skips future SQL optimizations.

— Names interface, package path and migration requirement clearly. v1.23.0
02 02
grpc-gateway v2.33.0 header and query parameter behavior change Not checked yet breaking 73

Upgrades to grpc-gateway v2.33.0, which no longer forwards Grpc-Metadata-* request headers to gRPCgRPCAn open-source remote procedure call framework from Google that uses HTTP/2 and Protocol Buffers to let services call each other's functions across a network, giving tools low-latency, strongly typed inter-process communication. metadata, and now rejects malformed query parameters (e.g. non-decimal float spellings, field[key] on non-map fields) that were previously accepted.

— Names dependency version and exact behavior changes affected users must check. v1.23.0
03 00
PostgreSQL latest-model lookup optimization Not checked yet improved 65

PostgreSQL's FindLatestAuthorizationModel now reads only the rows of the store's latest model instead of every model version, so Check and other requests made without authorization_model_id no longer slow down as model history grows.

— Clear mechanism and benefit but nothing for reader to configure. v1.23.0
└──▷ BREAKING ON UPGRADE
  • !
    Custom RelationshipTupleReader implementations (library use only) must now implement a Querier method on the interface (see pkg/storage/adapter); returning nil is safe but skips future SQL optimizations.
  • !
    grpc-gateway v2.33.0 (included in this release) no longer forwards Grpc-Metadata-* request headers to gRPC metadata, and now rejects malformed query parameters (e.g. non-decimal float spellings, field[key] on non-map fields) that were previously accepted.
Was this useful?

JumpCloud

Sources Blog / feed → Release Notes 2026 (2026-11-20) Blog Rank Why this rank top features scored 85/85/80, +10 product media, +6 runnable example — 99.33 How ranking works →

JumpCloud makes it simple to manage Windows, Apple, Linux, and Android devices and to deliver secure access with SSO, MFA, Zero Trust, RADIUS, and more.

JumpCloud's latest release headlines an audit-ready Access CertificationAccess CertificationA periodic review process in which managers or system owners confirm or revoke which users should retain access to systems and data, reducing privilege creep over time. workflow with multi-stage approval chains and signed exports, alongside new Agent Delegated Access for AI agents, capability-based MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. app setup, Workflow support in the MCP ServerMCP ServerA process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. for Admins, OAuth-enabled workflow connectors with an expanded template library, and enhancements to macOS patch policy enforcement, Device Lock, and SCIMSCIMA standard for creating, updating and removing user accounts and groups across applications. It helps an identity provider keep access in sync when people join, change roles or leave. group naming.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 6 FEATURES 2 OF 6 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
Access Certification review campaigns and exports Not checked yet new 85

JumpCloud adds an Access Certification feature supporting one-time or recurring review campaigns scoped to SSOSSOSingle sign-on, using one identity provider to sign in to multiple applications. It reduces separate passwords and centralizes account and access management. applications or directory user groups, with multi-stage approval chains and automated remediation that revokes access or suspends identities when deadlines expire.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Campaign snapshots can be exported in CSV, JSON, or PDF formats with SHA-256SHA-256A cryptographic hash function that turns data into a fixed 256-bit fingerprint. It is used to check integrity; hashing does not encrypt the original data. document hashing and formal administrator sign-off for compliance audits.

Launch a recurring quarterly access review campaign for an SSO application to satisfy compliance audits, then export the signed snapshot for your auditor.
📍In the Admin Portal, go to Access Certification, create a new campaign, select SSO Applications as the scope, set recurrence to quarterly, configure a multi-stage review chain (Managers then Resource Owners), enable automated remediation on deadline expiry, and after the review cycle completes export the campaign snapshot as PDF or CSV with administrator sign-off.
— Full mechanism, formats and hashing named, plus a walkthrough example Release Notes 2026 (2026-11-20)
02 02
Capability-based app setup for SSO, SCIM, MCP Not checked yet new 85

New applications only show configuration for the capabilities selected — SSO, SCIM, and/or MCP — with Manage MCP Servers as the selection for MCP-only access, letting admins create MCP-only applications, assign them to user groups for access control, and make the MCP server available through AI Gateway without configuring SSO or creating a bookmark. Admins can review a user group's applications from Identity Management > User Groups by opening the group and selecting the Applications tab, where apps with validated MCP configs are marked as ready for AI Gateway.

— Named config selection and exact navigation path given Release Notes 2026 (2026-11-20)
03 00
Workflow connector OAuth and expanded template library Not checked yet improved 80

Workflow connectors now support OAuth 2.0OAuth 2.0An open authorization framework (IETF RFC 6749) that lets a service grant limited API access to a third party without sharing credentials, giving cyber tools a standard way to authenticate against external platforms. authentication including Client CredentialsClient CredentialsAn OAuth 2.0 grant type where a service authenticates directly with an authorization server using its own ID and secret, with no user involved, to obtain an access token for machine-to-machine API calls. and Authorization Code with PKCEPKCEProof Key for Code Exchange, an OAuth safeguard that binds an authorization code to the client that requested it. An intercepted code cannot be redeemed without the client's matching verifier., with credentials stored in Secrets Manager and access tokensaccess tokensA credential that authorizes a client to use a service or API. It usually has limited permissions and a lifetime; anyone holding it may be able to use those permissions. refreshed automatically. The Workflow template library expands with ready-to-use templates: Auto-Suspend Users on Inactivity Alert, Auto-Block Newly Discovered Generative AI Apps, Mac Enrollment - Auto-Revoke Admin Privileges, Sensitive Software Install Detection - Email Notification, User Group Added - Send Member Email, User Activated - Add to Pre-Hire Holding Group If Before Start Date, and Agent Enrollment - Add Device Custom Attribute.

Block newly discovered generative AI apps automatically as they appear in your environment, using the new Workflow template.
📍In the Admin Portal, open the Workflow Builder template library, select Auto-Block Newly Discovered Generative AI Apps, customize the trigger and action settings for your environment, then save and activate the workflow.
— All templates and OAuth flows named with a usage walkthrough Release Notes 2026 (2026-11-20)
04 00
Device Lock command PIN escrow and status column Not checked yet improved 70

The Device Lock MDMMDMMobile device management, centrally configuring and managing enrolled devices. Administrators use it to apply settings, distribute applications and enforce device security policies. command for macOS, iOS, and iPadOS now supports automatic PIN generation and escrow for Mac (retrievable from Highlights > Device Lock > View PIN), a Lock Screen message and phone number, and a macOS Lock Status column in the Devices list.

— Exact navigation path and new list column named Release Notes 2026 (2026-11-20)
05 00
Separate enforcement for major and minor macOS updates Not checked yet improved 60

JumpCloud's DDM-Based macOS Patch Policy now supports separate enforcement settings for major OS upgrades and minor OS updates, including independent grace periods and enforcement times for each.

— Mechanism described but no UI path or config key given Release Notes 2026 (2026-11-20)
thinner coverage below
06 00
Workflow support in MCP Server for Admins Not checked yet new 55

The JumpCloud MCP Server for Admins adds Workflow support, letting admins create, modify, duplicate, troubleshoot, and manually trigger Workflows from a connected MCP-enabled AI client using natural language.

— Capability described but no concrete command or config shown Release Notes 2026 (2026-11-20)
Was this useful?

AlertEnterprise

Sources Blog / feed → Bring Guardian into the AI tools you already use (2026-11-20) Blog

AlertEnterprise is the future of work requires cyber-physical security convergence. Alert Enterprise provides identity & access SaaS solutions empowering businesses across the globe.

AlertEnterprise adds Guardian MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface., letting teams query and submit physical access and badge requests from AI assistants and developer tools under existing Guardian approvals.

  • New integrationNew Guardian MCP integration connects Guardian to AI assistants and developer tools so access, badge, and identity requests can be made directly from those tools under existing Guardian approvals. Not checked yet
  • Setup instructions for each supported AI tool are available at mcp.alertenterprise.io, covering both cloud-based and on-premise Guardian deployments. Not checked yet

2 details in Everything

Was this useful?

Auth0

Sources Blog / feed → Cross-App Access Requesting App now supports SAML in Early Access (2026-11-20) Blog

Secure users, AI agents, and more with Auth0, a production-ready authentication and authorization platform built to help you ship faster and drive sign-ups.

Auth0 adds SAMLSAMLA standard for passing signed identity assertions between a sign-in provider and an application. It lets a user sign in through their organization's identity provider. support to Cross-App Access Requesting App, enabling silent third-party API token fetch for SAML-authenticated enterprise users via Token Vault and ID-JAG.

  • Enable the Requesting Application role on a SAML connection and configure the issuer or discovery URL and client ID via the Dashboard or Management API so apps can silently fetch third-party API tokens for SAML-authenticated users without requiring manual account linking. Not checked yet
  • New integrationAuthenticate the Requesting App role with Private Key JWTJWTAn open standard (RFC 7519) for encoding claims as a signed, optionally encrypted JSON object passed between parties. Cyber tools use it to verify identity and permissions without querying a central session store. or a client secret when configuring the SAML connection in Cross-App Access. Not checked yet
  • In multi-org tenants, connection resolution in Cross-App Access is now organization-aware, rejecting requests when the enterprise connection or target resource app is not enabled for the specified org_id or when org context is ambiguous. Not checked yet

3 details in Everything

Was this useful?

Clerk

Sources Blog / feed → Protect your MCP server with Clerk (2026-11-20) Blog Rank Why this rank top features scored 35/30/26 — 30.33 How ranking works →

Clerk is an authentication and user-management platform that provides pre-built UI components, APIs, and SDKs for adding sign-up, sign-in, and identity features to applications.

Clerk introduced tooling to secure Model Context ProtocolModel Context ProtocolModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. (MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface.) servers, covering spec-compliant authorization, pre-auth tool access, and Server Card publishing.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 3 FEATURES 1 OF 3 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
MCP server authorization support Not checked yet new 35

Clerk added the ability to build MCP servers with authorization that follows the current MCP authorization spec.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Names the spec but no mechanism or API details Protect your MCP server with Clerk (2026-11-…
02 00
Pre-auth tool exposure for MCP Not checked yet new 30

Developers can open chosen MCP tools to users before sign-in.

— Describes capability but not configuration method Protect your MCP server with Clerk (2026-11-…
03 00
MCP Server Card publishing Not checked yet new 26

Clerk added support for publishing Server Cards for MCP servers.

— Bare feature name with no further detail Protect your MCP server with Clerk (2026-11-…
Was this useful?
◆  Data Security & DLP

bytebase

Sources Release notes → 3.25.0 Notes

Bytebase is a database governance built for humans and agents — controlling changes and access across every major database.

Bytebase adds AI SQL review (Enterprise), independent export permissions with a new SQL Editor Exporter role, MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. administer_workspace and request_access_grant tools, GitOpsGitOpsA practice where Git repositories serve as the single source of truth for infrastructure and application state, with automated systems continuously reconciling live environments to match what is declared in version control. release visibility, and the EXECUTE statement type for procedure calls.

└──▷ GET THIS VERSION
$ git clone --branch 3.25.0 https://github.com/bytebase/bytebase.git
# already have the repo? check out this version:
$ git checkout 3.25.0
  • AI SQL review comments on exact lines of a plan's SQL — flagging issues like an UPDATE without a WHERE — and blocks automatic rollout until all comment threads are resolved. Available on the Enterprise plan after configuring an AI provider. Not checked yet
  • New supported platformExport is now its own permission (bb.sql.export), separate from query access. A new SQL Editor Exporter role and an Allow downloading exports setting (Settings > General > SQL Editor) control whether and where exports can go. Not checked yet
  • New MCP tools administer_workspace (Read-write mode, user-confirmed) and request_access_grant extend what agents and Project Developer users can do over MCP. Not checked yet
  • GitOps Releases page now shows per-database file status, source commit, branch, author and pull request; bytebase-action 3.25.0 signs in via workload identityworkload identityAn identity assigned to software, such as a service or container, rather than a person. It lets the workload authenticate and receive permissions without sharing a user's credentials. and reports results in the job log. Not checked yet
  • Procedure calls and dynamic SQL are classified as EXECUTE with moderate default risk, enabling their own approval flow on every supporting engine. Not checked yet

5 details in Everything

└──▷ BREAKING ON UPGRADE
  • !The release detail page is retired; links now open the associated plan. Releases can no longer be archived or restored in the console.
  • !The Releases page now requires bb.plans.list instead of bb.releases.list.
  • !Listing a project's databases now requires bb.databases.list instead of bb.projects.get in custom roles.
  • !The Allow data export without approval setting is removed; custom roles with only bb.sql.select lose export until bb.sql.export is added.
  • !New SQL Editor User assignments no longer include export; SQL Editor Exporter must be assigned separately.
  • !Upgrading a workspace that required export approval turns Allow downloading exports off automatically.
  • !Approval rules matching procedure calls as unclassified statements must be updated to include EXECUTE.
  • !MCP tool inputs and outputs changed; dependent scripts and agent configurations must be reviewed.
Was this useful?

Zama fhEVM

Sources Release notes → v0.13.9 Notes Rank Why this rank top features scored 38 — 38 How ranking works →

FHEVM, a full-stack framework for integrating Fully Homomorphic Encryption (FHE) with blockchain applications

fhEVM's v0.13.9 patch backports key and parameter integrity checks to the SDK.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 1 FEATURE 0 OF 1 FEATURE most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
KMS generation key and CRS integrity checks Not checked yet improved 38

The SDK now adds KMSGeneration key and CRS integrity checks, backported to the v0.13.x line.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Names the check but no usage detail or mechanism given. v0.13.9
Was this useful?

MIND

Sources Press release → announced · Nov 19, 2026 Press Rank Why this rank top features scored 55 — 55 How ranking works →

MIND launched an Email DLPDLPData Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. module for Microsoft Exchange that classifies outbound messages and attachments and blocks inappropriate sends without installing an agent on Exchange.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

MIND data security platform puts data loss prevention (DLP) and insider risk management (IRM) on autopilot - automatically identify, detect & prevent data leaks

MIND launched an Email DLPDLPData Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. module for Microsoft Exchange that classifies outbound messages and attachments and blocks inappropriate sends without installing an agent on Exchange.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 1 FEATURE 1 OF 1 FEATURE most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
Email DLP module for Microsoft Exchange Not checked yet new 55

MIND adds an Email DLP module for Microsoft Exchange that classifies outbound messages and attachments and blocks inappropriate sends, with no agent installed on Exchange.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ The module shares the same policies as every other environment MIND protects (GenAI, Agentic AI, SaaSSaaSSoftware delivered over the internet and billed by subscription, hosted and operated by the vendor so the buyer skips installation and infrastructure management., on-premise file shares, and endpoints), extending outbound coverage to mobile devices.

— Explains mechanism and scope but no config/API specifics. launch-20261119-c0822155
Was this useful?
Governance, Risk & Compliance
◆  Continuous Control Monitoring (CCM)

RegScale

Sources Changelog → changelog · Nov 21, 2026 Changelog Rank Why this rank top features scored 90/85/85, +15 breaking change, +6 runnable example — 107.67 How ranking works →

RegScale is a continuous compliance management platform that automates security documentation, assessments, and regulatory compliance workflows for organizations.

RegScale adds a client_runs CLI command group to manage and unstick open sync runs, raises the default vulnerability-to-asset link cap to 50,000 with per-plugin Tenable matching, and surfaces Microsoft Defender recommendations as vulnerabilities, alongside a batch of JCAM CLI fixes, a Wiz inventory reclassification, and cleanup of unused scanner config blocks.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 10 FEATURES 0 OF 10 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Vulnerability-to-asset linking and Tenable plugin matching Not checked yet breaking 90

Scanner syncs now link a vulnerability to up to 50,000 affected assets by default (up from 10,000), configurable via maxAssetIdentifiersPerVulnerability in init.yaml; Nessus file import and Tenable.sc sync now record one vulnerability per Tenable plugin mapped to every affected asset instead of one per CVE, matching Tenable's own finding definition.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ On the first import after upgrading, earlier per-CVE vulnerabilities and their paired issues close and new per-plugin vulnerabilities open in their place.

Cap the per-vulnerability asset link count and restore per-finding grouping for a Trivy import that must stay compatible with an existing pipeline.
yaml
groupScanFindings: false
maxAssetIdentifiersPerVulnerability: 25000
— Named config key and numbers, with migration behaviour described. changelog-20261121-8737756e
02 00
client_runs command group for stuck syncs Not checked yet new 85

New regscale client_runs command group adds list, status, find, and close subcommands to inspect and unstick open sync runs directly from the CLI, so a stuck run can be found and closed before starting a new one.

Inspect and unstick a sync run that another process left open before starting a new one.
$ regscale client_runs list
regscale client_runs find --plan <plan-id> --source <source-name>
regscale client_runs close <run-id>
— Named subcommands plus a runnable example sequence. changelog-20261121-8737756e
03 00
JCAM CLI fixes across connection, mapping and export Not checked yet breaking 85

regscale jcam test_connection now exits 1 with a configuration message when jcamToken or jcamURL is unset or a placeholder instead of silently reporting success; import_parameters --learn-from now requires --crosswalk so a learned mapping is never written into the installed package where an upgrade would discard it; regscale jcam check_config now reports an unset jcamFrameworkCatalog catalogue ID for 800-53r5, which import_controls requires; regscale jcam export_ssp now sends each authorization and front-matter value as plain text from the mapped field, so drift detection no longer flags identical values and --force no longer fails with a serialization error; and export_poams now sends each POA&MPOA&MA Plan of Action and Milestones: a US federal government document that tracks known security weaknesses in a system, the remediation steps planned, and the deadlines for completing them, required under FISMA compliance.'s stored severity as the matching JCAM criticality instead of No Selection.

— Every JCAM subcommand and flag named with its exact behaviour change. changelog-20261121-8737756e
04 00
Wiz inventory Hardware-to-Software reclassification Not checked yet breaking 80

Wiz inventory sync now moves images stored as Hardware to Software and keeps their stored OS version; list VIRTUAL_MACHINE_IMAGE in wizHardwareAssetTypes with useWizHardwareAssetTypes on to retain them as Hardware. This takes effect on the next sync run after upgrading.

— Named config keys give a clear opt-out path. changelog-20261121-8737756e
05 00
Qualys VM import placeholder host ID keying Not checked yet breaking 70

Since RegScale 7.1.0, the Qualys VM scan import keys a host whose QG Host ID is a placeholder (none or unknown) on its IP address; a host imported earlier under a placeholder ID gets a new asset keyed on its IP, and its old asset stops receiving findings.

— Explains mechanism and consequence but no remediation step. changelog-20261121-8737756e
thinner coverage below
06 00
Defender recommendations as vulnerabilities Not checked yet new 55

Microsoft Defender and Defender for Cloud recommendations now also appear as vulnerabilities on the Vulnerability Status Board and in the FedRAMPFedRAMPA US government program that standardizes cloud security authorization, letting vendors get a single assessment accepted across federal agencies instead of repeating the process for each one. VDRVDRA Virtual Data Room is a secure, hosted document repository used for controlled sharing of sensitive files during due diligence, audits, or investigations, giving cyber tools a governed environment to exchange evidence or reports. report, linked to their existing issues; requires RegScale platform 6.36.2.0 or later.

— Clear scope and version requirement but no action step. changelog-20261121-8737756e
07 00
Removed no-op scanner config blocks Not checked yet deprecated 50

Removes no longer carries blocks for aqua, defenderFile, ecr, nexpose, snyk, sonarcloud, veracode, and xray from configuration, as the CLI never read them; existing blocks are now silently ignored.

— Names all affected keys but requires no reader action. changelog-20261121-8737756e
08 00
FedRAMP DRF import CVE handling fix Not checked yet improved 50

regscale fedramp import-drf no longer sets a deviation's CVE from the Vulnerability Name column when the workbook has no CVE column.

— Named command and column but minimal context. changelog-20261121-8737756e
09 00
Synqly filtered asset sync status fix Not checked yet improved 50

Synqly asset syncs run with --filter no longer mark assets excluded by the filter as Inactive.

— Named flag but no further mechanism given. changelog-20261121-8737756e
10 00
Tenable.io MAC address formatting Not checked yet improved 45

regscale tenable io sync_assets now sends each MAC addressMAC addressA hardware-address identifier used on a local network, such as Ethernet or Wi-Fi. It identifies a network interface for delivery on that link, rather than routing across the internet. in upper case split by colons.

— Named command, but a single-line formatting tweak. changelog-20261121-8737756e
└──▷ BREAKING ON UPGRADE
  • !
    Nessus and Tenable.sc now record one vulnerability per Tenable plugin instead of one per CVE; on the first import after upgrading, all earlier per-CVE vulnerabilities and their paired issues close, and new per-plugin vulnerabilities open in their place.
  • !
    JCAM import_parameters --learn-from now requires --crosswalk; calls without --crosswalk will fail.
  • !
    Wiz inventory sync will move images currently stored as Hardware to Software on the next run; set useWizHardwareAssetTypes and list VIRTUAL_MACHINE_IMAGE in wizHardwareAssetTypes to retain them as Hardware.
  • !
    Config blocks for aqua, defenderFile, ecr, nexpose, snyk, sonarcloud, veracode, and xray are no longer read by the CLI; existing blocks are silently ignored.
  • !
    Since 7.1.0, the Qualys VM scan import keys a host whose QG Host ID is a placeholder (none or unknown) on its IP address, so a host imported earlier under a placeholder ID gets a new asset keyed on its IP and its old asset stops receiving findings.
Was this useful?
◆  Integrated & IT Risk Management

CISO Assistant

Sources Release notes → v4.3.4 Notes Rank Why this rank top features scored 65/50/50, +15 breaking change — 70 How ranking works →

CISO Assistant v4.3.4 adds a GLPIGLPIOpen-source IT asset management and helpdesk platform maintained by the Teclib community; cyber tools integrate with it to correlate vulnerability findings against a live inventory of hardware, software, and users. integration, an experimental pre-audit preparation module, new findings-to-assessment linking, TPRMTPRMAssessing and monitoring the security risk an organization takes on through its vendors, suppliers and service providers. It covers the questionnaires, evidence and continuous checks behind letting a third party hold your data. task assignment, several new compliance frameworks and mappings, and warns that curl will be removed from the next release's Docker image.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

CISO Assistant is a one-stop-shop GRC platform for Risk Management, AppSec, Compliance & Audit, TPRM, BIA, Privacy, and Reporting. It supports 200+ global frameworks with automatic control mapping, including ISO 27001, NIST CSF, SOC 2, CIS, PCI DSS, NIS2, DORA, GDPR, HIPAA, CMMC, and more.

CISO Assistant v4.3.4 adds a GLPIGLPIOpen-source IT asset management and helpdesk platform maintained by the Teclib community; cyber tools integrate with it to correlate vulnerability findings against a live inventory of hardware, software, and users. integration, an experimental pre-audit preparation module, new findings-to-assessment linking, TPRMTPRMAssessing and monitoring the security risk an organization takes on through its vendors, suppliers and service providers. It covers the questionnaires, evidence and continuous checks behind letting a third party hold your data. task assignment, several new compliance frameworks and mappings, and warns that curl will be removed from the next release's Docker image.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 9 FEATURES 1 OF 9 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
New compliance frameworks and cross-framework mappings Not checked yet new 65

Adds the SAMA IT Governance Framework, the NDMO Data Management and Personal Data Protection Standards (SDAIA), the CADA and SEAL frameworks for the European market, and Essential EightEssential EightA set of eight baseline mitigation strategies published by the Australian Signals Directorate to help organisations prioritise defences against the most common cyber threats. reference controls, threats, and mappings to the library.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Also adds CMMC to NIST 800-171 rev2, NIST 800-171 rev2 to rev3, and CMMC to CSF 2.0 mappings to the library.

— Names every framework and mapping added, clear library location v4.3.4
thinner coverage below
02 00
Opt-in auto-lock for overdue audits Not checked yet improved 50

Auto-lock of overdue audits is now controlled by a new setting and is off by default, whereas it was previously automatic.

— Explains before/after behaviour but not exact setting key v4.3.4
03 00
curl removed from Docker health check Not checked yet breaking 50

curl will be removed in the next version — users should update their Docker Compose health check before upgrading.

— Gives a clear pre-upgrade action but limited mechanism detail v4.3.4
04 00
GLPI integration for tickets and assets Not checked yet new 40

Adds a GLPI provider integration for ticket and asset management.

— Names the integration but no setup or endpoint detail v4.3.4
05 00
Task assignment to third-party audit representatives Not checked yet new 40

In the TPRM module, tasks raised on a third-party audit can now be assigned to its representatives.

— Names module and action but no UI path v4.3.4
06 00
Audit reports show implementation, documentation, maturity scores Not checked yet improved 40

Audit reports now display implementation, documentation, and maturity scores.

— Names the three new score types shown v4.3.4
07 00
Attach findings to requirement assessments Not checked yet new 35

Findings can now be attached directly to a requirement assessment.

— States the capability without workflow detail v4.3.4
08 00
German and Hungarian localization additions Not checked yet new 35

Adds German translations for CRA, metrics, workflow, and threats to the library, and adds Hungarian email templates and policy templates.

— Names languages and content areas but no usage detail v4.3.4
09 00
Experimental pre-audit preparation module Not checked yet new 25

Adds an experimental pre-audit preparation module.

— Bare name, no mechanism or scope given v4.3.4
└──▷ BREAKING ON UPGRADE
  • !
    curl will be removed in the next version — update your Docker Compose health check before upgrading.
Was this useful?
◆  Integrated & IT Risk Management

raviaxo RiskShard

Sources Release notes → v0.15.0 Notes Rank Why this rank top features scored 70/60 — 65 How ranking works →

RiskShard added evidence re-verification tracking and expanded its measurement basis vocabulary with a geometric-mean cost option.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

RiskShard is an evidence-governed quantitative cyber risk — a trustworthy CLI and scenario engine where every number traces to a reviewed public source.

RiskShard added evidence re-verification tracking and expanded its measurement basis vocabulary with a geometric-mean cost option.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 2 FEATURES 0 OF 2 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Evidence re-verification fields for artifacts Not checked yet new 70

New artifact_verified_at and artifact_verified_sha256 fields track when evidence was re-read against its source file and detect publisher reissues automatically; the two fields must appear together or not at all.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Names exact fields and their pairing rule. v0.15.0
02 00
Geometric-mean cost measurement basis Not checked yet new 60

A new geometric_mean_total_event_cost value models geometric-mean cost figures from loss reports, added as the 19th entry in the measurement_basis vocabulary.

— Names the exact value and its position in vocabulary. v0.15.0
Was this useful?
◆  Compliance Automation & Trust Management

Compyl

Sources Release page → Nov 19th, 2026 Notes

Compyl is the AI-powered GRC platform that maps one control library to SOC 2, ISO 27001, HIPAA, PCI DSS and 70+ frameworks, so you stay audit-ready year-round.

Compyl adds custom Questionnaire fields queryable in Query Builder, bulk blueprint scheduling, and a Trust Center Admin path change to /trust-center-admin.

  • Add custom fields to Questionnaires: admins can create an extension form whose fields appear on each questionnaire, save per-response data, and are queryable in Query Builder. Not checked yet
  • Set a bulk run schedule for blueprints to move many blueprints to a new cadence (for example, monthly to daily) in one step. Not checked yet
  • List fields such as the VirusTotal watch list now include search, a clear-all action, and an action to delete all entries that failed validation. Not checked yet
  • The Trust Center admin module now lives at /trust-center-admin and is renamed Trust Center Admin in navigation, freeing /trust-center for the public Trust Center page. Not checked yet

4 details in Everything

└──▷ BREAKING ON UPGRADE
  • !The Trust Center admin module has moved from its previous path to /trust-center-admin; any bookmarks or hardcoded links to the old admin URL must be updated.
Was this useful?
Agent
◆  MCP Servers & Tooling

Runlayer

Sources Blog / feed → 6.5.0 (2026-11-21) Blog Rank Why this rank top features scored 75/75/70, +6 runnable example — 79.33 How ranking works →

Runlayer added PGPPGPA public-key encryption standard for signing and encrypting data, originally by Phil Zimmermann; cyber tools use it to verify software authenticity and protect sensitive communications. private key detection and audit-payload credential masking, Cursor blocking notices, and faster PIIPIIPersonally Identifiable Information — data that can identify a specific individual, such as names, email addresses, or SSNs. Cyber tools flag or redact it to limit exposure and meet privacy regulations./token masking at scale, alongside a new access-request workflow, reply-to-quote chat, and a unified Procedures setting that auto-benchmarks and auto-promotes self-improving agents' procedure drafts.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Runlayer is an MCP gateway that enforces identity-mapped permissions and detects threats across the agents and MCP servers employees use.

Runlayer added PGPPGPA public-key encryption standard for signing and encrypting data, originally by Phil Zimmermann; cyber tools use it to verify software authenticity and protect sensitive communications. private key detection and audit-payload credential masking, Cursor blocking notices, and faster PIIPIIPersonally Identifiable Information — data that can identify a specific individual, such as names, email addresses, or SSNs. Cyber tools flag or redact it to limit exposure and meet privacy regulations./token masking at scale, alongside a new access-request workflow, reply-to-quote chat, and a unified Procedures setting that auto-benchmarks and auto-promotes self-improving agents' procedure drafts.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 9 FEATURES 0 OF 9 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Access request workflow for artifacts and agents Not checked yet new 75

Users can request access to any artifact, agent, skill, or plugin they lack permission to open by clicking Request access; the owner receives the request in Approvals and in Slack, and approving it opens the page for the requester.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Grant a teammate access to an agent they cannot open, without leaving the product, by using the built-in request flow.
📍In the Runlayer web app, open the agent, artifact, skill, or plugin page you cannot access, then click Request access. The owner sees the request under Approvals (and in Slack if connected) and can share it directly from there.
— Includes exact UI steps and notification channels 6.4.0 (2026-11-20)
02 00
Procedures setting for self-improving agents Not checked yet new 75

A single Procedures setting lets self-improving agents save their own procedures, automatically triggers benchmarking of drafts saved without named runs, and promotes a passing draft to live without waiting for approval. Stale procedure draft versions with no review decision after 30 days are automatically sent back with a reason and their review requests closed, while any live version keeps running; failed procedure benchmarks now retry daily automatically instead of waiting for a new version, within each version's test-run limit.

— Full lifecycle described but no config key named 6.4.0 (2026-11-20)
03 00
PGP private key detection in tool-call arguments Not checked yet new 70

Runlayer now detects complete armored PGP private keys (-----BEGIN PGP PRIVATE KEY BLOCK-----, PGP SECRET KEY BLOCK) and private keys with dash-containing headers (GPGTools Comment: line, encrypted PEM DEK-Info:) in tool-call arguments, so the Mask, Block and Alert actions now apply to them.

— Names exact detected patterns but no config surface given 6.5.0 (2026-11-21)
04 00
MCP output blocking notices in Cursor Not checked yet improved 70

On Cursor 2.6.22 or later, Runlayer now shows a notice to the agent when an MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tool's output is blocked in Protect mode (replacing the previous silent omission) and also shows a notice when Runlayer cannot be reached to scan an MCP tool's output. Sessions and the supported clients list now report Block bypassed for built-in tool output and Blocked for replaced MCP output.

— Clear before/after and version gate, no config key 6.5.0 (2026-11-21)
05 00
Reply-to-quote in agent chat Not checked yet new 70

Highlighting words in an agent's answer and clicking Reply quotes them into your message box; available in agent chat, the Runlayer Assistant, and artifact chat.

Quote a specific sentence from an agent response to ask a follow-up question without retyping the context.
📍In agent chat, highlight the relevant words in the agent's answer, then click Reply — the quoted text appears in your message box and is sent with your next message.
— Exact UI interaction given, easy to try 6.4.0 (2026-11-20)
06 00
Token masking improvements Not checked yet improved 60

Hidden-character, PII and token-masking scanner time now scales proportionally to tool-call or tool-result size instead of slowing down sharply on dense inputs, so a multi-megabyte response with many personal-data matches that previously took minutes to mask now takes about a second; token masking also now fully hides a private key written right after password= with no END marker, where previously only the first few characters were masked.

— Quantifies speedup but names no config to tune it Specific before/after but no config surface 6.5.0 (2026-11-21)6.4.0 (2026-11-20)
07 00
Skill risk policy blocks extended to @ mentions and distillation Not checked yet improved 60

Skill risk policy blocks now also apply when skills are reached via @ mentions in agent chat or through skill distillation in self-improving agents, matching the behaviour already in place for agent runs.

— Names exact mechanism but no config surface 6.4.0 (2026-11-20)
thinner coverage below
08 00
Session privacy in Analytics detail view Not checked yet improved 55

When Session privacy is on, the timeline in an Analytics output's detail view now hides other people's sessions, showing a count of hidden sessions instead of their title, owner, cost or session link.

— Describes behaviour but no setting path shown 6.5.0 (2026-11-21)
09 00
Credential masking in audit payloads Not checked yet new 55

Recognised credentials are now masked in audit payloads before they reach the audit database, payload archives and downstream exports; security approvals retain matching without copying detected credential values into approval evidence.

— Explains pipeline stages but no API or flag named 6.5.0 (2026-11-21)
Was this useful?

mcp-security

Sources Release notes → secops-soar-v0.2.0 Notes Rank Why this rank top features scored 95/65/55, +15 breaking change, +6 runnable example — 92.67 How ranking works →

mcp-security shipped new case-management tools for its SOARSOARSecurity orchestration, automation and response, coordinating incident workflows across tools. Playbooks can gather evidence, assign work or carry out approved response actions. server, added proxy support across both the SOAR and GTI (VirusTotal) clients, and introduced a breaking configuration gate on GTI's file-analysis tool.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

mcp-security shipped new case-management tools for its SOARSOARSecurity orchestration, automation and response, coordinating incident workflows across tools. Playbooks can gather evidence, assign work or carry out approved response actions. server, added proxy support across both the SOAR and GTI (VirusTotal) clients, and introduced a breaking configuration gate on GTI's file-analysis tool.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 4 FEATURES 1 OF 4 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
GTI_ALLOWED_UPLOAD_DIR required for file analysis Not checked yet breaking 95

The analyse_file tool now requires the GTI_ALLOWED_UPLOAD_DIR environment variable to be configured before it is enabled, restricting which directories may be uploaded; existing deployments relying on analyse_file without this setting will have file analysis disabled until the variable is set, e.g.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ GTI_ALLOWED_UPLOAD_DIR=/tmp/gti-uploads.

Allow the analyse_file tool to upload files from a specific directory by setting GTI_ALLOWED_UPLOAD_DIR in your environment before starting the server.
env
GTI_ALLOWED_UPLOAD_DIR=/tmp/gti-uploads
— Exact env var, breaking impact, and example command given. gti-v0.1.4
02 00
Proxy support via trust_env in SOAR and GTI clients Not checked yet new 65

Corporate proxy support was added by setting trust_env in the HttpClient session used by the SOAR MCP server, and by setting trust_env during vt.Client initialization in the GTI VirusTotal client, so HTTP proxy environment variables are honoured automatically in both.

— Mechanism and client names given, but no exact env var shown. gti-v0.1.4secops-soar-v0.2.0
thinner coverage below
03 00
Case management tools in SOAR MCP server Not checked yet new 55

The SOAR MCP server adds create_case and close_case tools so cases can be created and closed programmatically.

— Names new tools but gives no usage example. secops-soar-v0.2.0
04 00
Identifier validation and path normalization in HttpClient Not checked yet improved 40

HttpClient now validates case and alert identifier formats and normalizes request paths automatically, reducing malformed requests to the SOAR API.

— Internal behavior change with no exposed config or command. secops-soar-v0.2.0
└──▷ BREAKING ON UPGRADE
  • !
    The analyse_file tool now requires GTI_ALLOWED_UPLOAD_DIR to be configured; existing deployments that relied on analyse_file without this setting will have file analysis disabled until the variable is set.
Was this useful?
Govern
◆  Model & Agent Protection

ai-safe2-framework

Sources Release notes → v3.5 Notes Rank Why this rank top features scored 88/85/75 — 82.67 How ranking works →

AI SAFE² v3.5 adds three CLI routes for querying, auditing, and assessing readiness against framework controls, an offline Rosetta Stone explorer consolidating over a thousand source-to-control relationships, and twelve new governance controls alongside expanded MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface., deployment, UAS, and threat profiles.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Ai-Safe2-Framework provides governance, risk, and compliance controls for securing agentic AI systems and non-human identities.

AI SAFE² v3.5 adds three CLI routes for querying, auditing, and assessing readiness against framework controls, an offline Rosetta Stone explorer consolidating over a thousand source-to-control relationships, and twelve new governance controls alongside expanded MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface., deployment, UAS, and threat profiles.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 9 FEATURES 0 OF 9 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
New CLI routes for framework query, audit, readiness Not checked yet new 88

Adds three new CLI routes: safe2 framework query --control CP.12 to query a specific control, safe2 framework audit to audit against framework controls, and safe2 framework readiness --help to assess readiness.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

— Exact runnable commands with flags given. v3.5
02 00
Offline Rosetta Stone explorer and consolidated source data Not checked yet new 85

Open the offline Rosetta Stone explorer by loading dashboard/rosetta.html from a local checkout to filter records by source, control, review status, and keyword, and export them as JSON. It now consolidates 1,202 recovered source-to-control relationships across 33 external source families, 64 discovery entries, and 16 audit-rule classes.

— Exact file path and filter/export mechanics given. v3.5
03 00
12 new controls across trust and evidence domains Not checked yet new 75

Adds 12 new controls — CP.11 through CP.15, S1.8, F3.6, F3.7, A2.7, A2.8, A2.9, and E5.5 — covering trust lifecycle governance, final-effect authorization, control evidence independence, organizational AI management governance, design-time coordination contracts, pre-trust initialization boundaries, aggregate consumable-authority ceilings, environment binding, idempotency and side-effect ledgers, verified outcome contracts, evidence-based failure attribution, and declared success criteria gates.

— Each control ID named with its specific governance scope. v3.5
thinner coverage below
04 00
UAS v2 revises acceptance interpretation Not checked yet improved 55

Adds UAS v2, retaining 27 IDs while revising acceptance interpretation for human understanding, appropriate reliance, intervention, fairness, and evidence quality, with reciprocal Cognitive Sovereignty Framework references.

— Scope and ID count given, no lookup path. v3.5
05 00
157 child clauses and research notes deepen existing controls Not checked yet improved 50

Adds 157 explicit child clauses across 53 reviewed decisions, including two research-disposition clauses, plus 53 new research notes connecting source material, placement decisions, alternatives, trade-offs, and acceptance cases to framework controls — deepening requirement precision across existing controls.

— Counts given but no named controls or access point. v3.5
06 00
Six deployment profiles for execution surfaces Not checked yet new 50

Adds six deployment profiles covering harnesses, gateways, evaluation, browser and computer use, finance, and physical systems to make applicability explicit across execution surfaces.

— Lists surfaces but no profile IDs or access path. v3.5
07 00
Three new threat profiles sharpen existing threat families Not checked yet new 40

Adds three threat profiles sharpening attention to assurance compromise, unsafe authorized execution, and human-agent trust exploitation within the existing eleven threat families.

— Named threat areas but no control mapping shown. v3.5
08 00
MCP profile expanded with 12 child requirements Not checked yet improved 35

Expands the MCP profile with 12 additional child requirements across the existing 19 stable control IDs.

— Numbers given but no specific requirement names. v3.5
09 00
Machine-readable page map and agent discovery indexes Not checked yet new 35

Adds a machine-readable complete page map, root and NEXUS agent indexes, and shared module navigation to support both human and agent repository discovery.

— Describes purpose but no file paths or formats named. v3.5
Was this useful?

Minimal Software Research minimal

Sources Release notes → release-14b57239 Notes Rank Why this rank top features scored 80/75/50, +6 runnable example — 74.33 How ranking works →

minimal's release-14b57239 deepens its min box CLI with audit, inspection and egress controls, and lays groundwork for secrets management, hardware-backed key storage and OpenTelemetryOpenTelemetryAn open standard and set of tools for collecting traces, metrics and logs from applications. It lets services export observability data without tying instrumentation to one monitoring vendor. tracing, though most of the latter are named without further detail.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Minimal is a secure, declarative runtimes for Claude Code, Gemini, Codex, Pi and more, so you can run multiple agents safely on your stack from a verified software supply chain.

minimal's release-14b57239 deepens its min box CLI with audit, inspection and egress controls, and lays groundwork for secrets management, hardware-backed key storage and OpenTelemetryOpenTelemetryAn open standard and set of tools for collecting traces, metrics and logs from applications. It lets services export observability data without tying instrumentation to one monitoring vendor. tracing, though most of the latter are named without further detail.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 9 FEATURES 0 OF 9 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 00
Box inspection and introspection commands Not checked yet new 80

min box show <box> --network shows a box's effective network policy; min box show -o json and min box show -o yaml output box details including its exec list; min box list/show subcommands list and show box types; min box spec lists each store reference with its authorities and each lifecycle hook with its declarer; and box names are shell-completable for every min box verb that accepts one.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Inspect the effective network policy of a running box to confirm egress rules before an agent run.
$ min box show mybox --network
— Multiple named subcommands and flags enumerated with a usage example. release-14b57239
02 00
Box audit trail with parent history merge Not checked yet new 75

Read a box's audit trail with min box audit, and merge parent history into the trail with min box audit --parent to verify no decisions were edited.

Review the full audit trail for a box, merging parent history, to verify no decisions were edited.
$ min box audit mybox --parent
— Named flag with a runnable verification command. release-14b57239
thinner coverage below
03 00
Box Egress Proxy and egress flags on run commands Not checked yet new 50

minimal adds a Box Egress Proxy for controlling network egress from boxes, alongside new network and egress flags on min task run and min box run.

— Named flags and a proxy feature but no mechanism or example given. release-14b57239
04 00
Vulnerability reporting via min diag --security Not checked yet new 50

min diag --security reports a suspected vulnerability.

— Named flag with clear purpose but no detail on vulnerability scope. release-14b57239
05 00
Tampered audit chain detection in min doctor Not checked yet new 40

min doctor now reports an edited audit chain, flagging boxes whose audit history has been altered.

— Names the command and detection purpose, no example shown. release-14b57239
06 00
Software key names shown in min auth status Not checked yet improved 40

min auth status now shows software key names for each key.

— Named command enhancement without a usage example. release-14b57239
07 00
Secrets management via min secret subcommands Not checked yet new 30

minimal introduces secrets management with new min secret subcommands.

— Subcommand prefix named but no specific options described. release-14b57239
08 00
TPM/Secure Enclave key storage Not checked yet new 20

minimal adds support for storing keys in a TPMTPMA dedicated hardware chip, standardized by the Trusted Computing Group, that stores cryptographic keys and measurements in tamper-resistant silicon, giving cyber tools a root of trust that software alone cannot fake. or Secure Enclave.

— Only named in the release summary, no usage detail. release-14b57239
09 00
OpenTelemetry tracing support Not checked yet new 20

minimal adds OpenTelemetry tracing.

— Named only, no configuration or scope detail provided. release-14b57239
Was this useful?

ToolHive

Sources Release notes → v0.67.0 Notes Rank Why this rank top features scored 90/85/75, +15 breaking change, +6 runnable example — 104.33 How ranking works →

ToolHive is an enterprise-grade platform for running and managing Model Context Protocol (MCP) servers.

ToolHive v0.67.0 adds a client_credentialsclient_credentialsAn OAuth 2.0 grant type where a service authenticates directly with an authorization server using its own ID and secret, with no user involved, to obtain an access token for machine-to-machine API calls. grant for SPIFFESPIFFEAn open standard, maintained by the CNCF, for assigning cryptographic identities to workloads in dynamic infrastructure so services can authenticate each other without passwords or static credentials.-authenticated workloads with per-association grant-type selection, structured severity findings for skill and plugin validation, and a configurable DCR upstream client name — alongside several breaking changes to SPIFFE token exchangetoken exchangeAn OAuth 2.0 protocol extension (RFC 8693) that lets a service trade one token for another with different scopes, audience, or type, enabling delegation and impersonation across service boundaries without re-authenticating the user., default transport selection, and MCPRemoteProxy resource limits.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

└──▷ WHAT SHIPPED · 6 FEATURES 3 OF 6 FEATURES most completely described first
what's the number?

Each feature carries 0–100 for how completely the vendor documented it — not how big or important the work is. A major capability described in eight words scores low, and that is the finding.

  • depth0–40 what it does and how it works or what changed
  • specificity0–30 names real surfaces — APIs, flags, formats, limits, numbers
  • actionability0–30 enough to go use it — a named endpoint, flag, or config key tops this; a UI path is a starting point

Code and config rank first by construction: a documented endpoint, flag or config key scores at the top of specificity and actionability, so it sorts above a feature described only in prose. Under 60 the notes go thin — everything below the “thinner coverage” line is thinner documentation, not smaller work. Hover any meter for that feature's three sub-scores.

01 01
SPIFFE token exchange breaking changes Not checked yet breaking 90

SPIFFE associations (spiffeClientAuth) now require a non-empty resources list; associations without resources are rejected and the auth server refuses to start, and the audiences field is now ignored.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Token-exchange requests must send exactly one resource parameter — requests sending audience= or omitting a target now receive invalid_target. The act.sub claim in a SPIFFE token exchange is now the workload's SPIFFE ID instead of the client ID, so Cedar policies matching claim_act.sub against a SPIFFE client ID must be updated to match claim_client_id. The Go API's token-exchange HTTP middleware and runner factory must now be imported from pkg/auth/tokenexchange, and authserver.SPIFFEGrantTypeTokenExchange must be replaced with oauthproto.GrantTypeTokenExchange.

— Enumerates every migration step needed v0.67.0
02 00
SPIFFE client_credentials grant and grant-type selection Not checked yet new 85

MCPExternalAuthConfig and VirtualMCPServer SPIFFE associations now have a grantTypes field to select client_credentials, token exchange (urn:ietf:params:oauth:grant-type:token-exchange), or both, defaulting to token exchange. SPIFFE-authenticated workloads can now obtain access tokensaccess tokensA credential that authorizes a client to use a service or API. It usually has limited permissions and a lifetime; anyone holding it may be able to use those permissions. via the client_credentials grant bound to their SPIFFE ID, with token lifetime capped at the SVID's remaining validity.

Allow a SPIFFE workload to authenticate with both client_credentials and token exchange on a VirtualMCPServer association.
yaml
grantTypes:
  - client_credentials
  - urn:ietf:params:oauth:grant-type:token-exchange
— Names config field, grant types, and validity rule v0.67.0
03 02
Default resource limits for MCPRemoteProxy Not checked yet breaking 75

MCPRemoteProxy now applies default CPUCPUThe main general-purpose processor in a computer. It runs application instructions and coordinates work with memory and other hardware. (50m request / 200m limit) and memory (64Mi request / 256Mi limit) resource constraints; proxies whose working set exceeds 256Mi will be OOMKilled on upgrade unless explicit limits are set beforehand.

— States exact default limits and upgrade risk v0.67.0
04 03
Default transport change for uvx packages Not checked yet breaking 70

thv run uvx://… now defaults to stdio transport instead of auto-detecting; packages that serve HTTP or SSESSEServer-Sent Events, a browser API and HTTP-based protocol where a server pushes a stream of text events to a client over a single long-lived connection, giving tools a lightweight alternative to WebSockets for one-way real-time updates. must now explicitly pass --transport streamable-http or --transport sse.

— Gives exact flags needed to migrate v0.67.0
05 00
DCR upstream client name configuration Not checked yet new 65

Operators can set clientName in the DCR upstream configuration to control the name shown on the upstream IdPIdPA service that authenticates users and supplies identity information to other applications. It is the sign-in authority behind many single sign-on systems.'s authorization screen; it defaults to ToolHive MCP Client.

Override the client name shown on the upstream IdP consent screen when using DCR upstream configuration.
yaml
clientName: "My Team MCP Client"
— Names config key and default, has example v0.67.0
06 00
Structured findings in skill and plugin validation Not checked yet new 65

thv skill validate, thv ai-plugin validate and their corresponding API endpoints now return a structured findings array with severity levels instead of plain text output.

— Names commands and response field but no example v0.67.0
└──▷ BREAKING ON UPGRADE
  • !
    SPIFFE associations (spiffeClientAuth) now require a non-empty resources list; associations without resources are rejected and the auth server refuses to start. The audiences field is now ignored.
  • !
    SPIFFE token-exchange requests must send exactly one resource parameter; requests that send audience= or omit a target now receive invalid_target.
  • !
    thv run uvx://… now defaults to stdio transport; packages that serve HTTP or SSE must now explicitly pass --transport streamable-http or --transport sse.
  • !
    The act.sub claim in a SPIFFE token exchange is now the workload's SPIFFE ID instead of the client ID; Cedar policies matching claim_act.sub against a SPIFFE client ID must be updated to match claim_client_id.
  • !
    MCPRemoteProxy now applies default CPU (50m request / 200m limit) and memory (64Mi request / 256Mi limit) resource constraints; proxies whose working set exceeds 256Mi will be OOMKilled on upgrade unless explicit limits are set beforehand.
  • !
    Go API: the token-exchange HTTP middleware and runner factory must now be imported from pkg/auth/tokenexchange; authserver.SPIFFEGrantTypeTokenExchange must be replaced with oauthproto.GrantTypeTokenExchange.
Was this useful?
Saved in this browser. Sign in to sync your tools.