The Cyber Toolchain — issue 072, September 12, 2026

Continuous Scanning triggers a full scan on every meaningful commit to a default branch, including merged pull requests, with follow-up scans closing findings fixed by those commits.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ It's configurable per organization, tag, or repository on GitHub, GitLab, Bitbucket Cloud, and Bitbucket Server.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Repository objects now expose lastScanFailure, replacing lastScanError, carrying both the error message and the outcome. Any integration or script reading lastScanError will receive no value after upgrade.

Approved patches now show pull request submission status — queued, in-progress step, retry state, or stop reason — with Cancel and Retry actions and up to five automatic submission attempts. A completed pull request scan links to the findings list it produced, with a separate dependency findings link.

ZeroPath imports CVS modules by server path over SSH or pserver for full scans, scheduled scans, dependency analysis, and secrets scanning.

CSV exports and the export endpoint now include Jira and Linear ticket links. An on-demand Jira ticket sync preview shows each linked ticket's pending action before you apply it. Ticket creation for Jira and Linear is now a per-repository permission requiring an explicit grant when access control is enabled.

ZeroPath imports Gerrit projects for full scans and scans each open change as new patch sets arrive, posting findings as a review message on the patch set.

Assistant skills are reusable methodology definitions — triage, dependency reachability, report writing, patch review — that the Assistant applies while working. They're authorable from the Skills tab, with four built-in forkable skills.

A finding ZeroPath has not assessed now appears in the queue as Needs Review, ranked on severity alone — widening the previous rule that only verified findings appeared, so open counts will rise after upgrade.

Project HydraFusion is now available under /experimental in Copilot CLI, providing automated semantic routing between local, cloud, and compound models.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Select it like any other model and it balances performance, cost, and latency per task.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

⤢

The Copilot usage metrics API adds VS Code Agents fields: daily_active_vscode_agent_users (unique users active in the VS Code Agents window per 1-day/28-day period) and totals_by_vscode_agent (optional session_count and total_user_messages rollups) in aggregate enterprise/organization reports, plus used_vscode_agent and per-user totals_by_vscode_agent in per-user reports.

Public preview of centralized sandbox policy controls for GitHub Copilot in JetBrains IDEs, letting enterprise administrators govern sandbox enablement, filesystem and network access, proxy settings, developer-tool access, and macOS Keychain access.

Adds a --llm-mode CLI flag and mode field in the [LLM] config-file section to switch Black Duck Signal between Black Duck-managed and BYO-LLM endpoints.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ The global [LLM] section sets uri, key, timeout, gcp_service_account_key, and gcp_project defaults for all seven agents, while per-agent sections — [LLM_SINGLEFILESCAN_AGENT], [LLM_DATAFLOW_AGENT], [LLM_OVERSIGHT_AGENT], [LLM_EXPLOITATION_AGENT], [LLM_SCA_AGENT], [LLM_MERGEKEY_EMBED_AGENT], [LLM_OVERSIGHT_EMBED_AGENT] — accept model, uri, key, timeout, and api_version overrides so each agent (single-file SASTSASTStatic application security testing, analyzing source code or compiled code for vulnerabilities without running the application. It helps developers find problems before deployment., cross-file dataflow, oversight, exploitation, SCASCASoftware composition analysis, identifying third-party dependencies and checking them for known vulnerabilities and license issues. It covers software a team uses rather than writes itself., and two embeddingembeddingLists of numbers a model produces to represent the meaning of text, images or code, so similar items sit close together. They power semantic search and RAG. agents) can target a different provider, model, and endpoint within one scan.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Supported BYO-LLM providers are OpenAI direct API, Azure OpenAI (including Azure AI Services with native Anthropic Claude and automatic Claude-vs-OpenAI family detection), AWS Bedrock via 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. bearer-token auth, GCP Vertex AI with native service-account credentials and automatic token refresh, and any OpenAI-compatible proxy including LiteLLM.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

[LLM]
mode = byo
uri = https://your-endpoint.example.com/v1
key = ${OPENAI_API_KEY}
timeout = 600

[LLM_SINGLEFILESCAN_AGENT]
model = gpt-4o
timeout = 600

[LLM_DATAFLOW_AGENT]
model = gpt-4o-mini

[LLM_OVERSIGHT_AGENT]
model = gpt-4o-mini

[LLM_EXPLOITATION_AGENT]
model = gpt-4o-mini

[LLM_SCA_AGENT]
model = gpt-4o-mini

[LLM_MERGEKEY_EMBED_AGENT]
model = text-embedding-3-large

[LLM_OVERSIGHT_EMBED_AGENT]
model = text-embedding-3-small

[LLM]
mode = byo
gcp_service_account_key = gcp-sa-key.json
gcp_project = your-gcp-project-id

Adds a --check-llm preflight flag that validates every configured LLM endpoint is reachable and authenticated before a scan runs, producing machine-readable JSON output for CI pipelines.

Adds the veracode fix sca CLI command to analyze vulnerable dependencies surfaced by SCA scans, generate remediation patches, and identify safe upgrade paths.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Currently Limited Availability.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

$ veracode fix sca

The Get Lesson Progress API adds an updatedAfter parameter for incremental loads, an excludeNotStarted=true parameter to filter unstarted lessons, and new stack and updated_on fields in the response; it now returns a Not Started status when both start_time and end_time are null.

Adds full OWASPOWASPA nonprofit foundation that publishes open standards, guides, and ranked vulnerability lists—most famously the OWASP Top 10—giving cyber tools a vendor-neutral benchmark to test or report against. Top 10:2025 Security Labs curriculum including two new categories, A03 Software Supply Chain Failures and A10 Mishandling of Exceptional Conditions, plus new eLearning courses aligned to OWASP Top 10:2025 for .NET, Java, JavaScript, PHP, Python, and general Software Security Awareness, available with English, Spanish, French, Portuguese, and Chinese captions.

The Security Labs default URL has changed from https://securitylabs.veracode.com/ to https://securitylabs.veracode.com/security-labs/; customers using SSO must update their configured Security Labs URL.

Removes 33 deprecated eLearning courses (OWASP 2017, OWASP Top 10 (2021), and Secure Coding for .NET/.Java/.JavaScript/.PHP/.Python, 6 courses each except 3 for Python) and the OWASP Top 10 (2021) Security Labs lessons; certification levels have been updated to include newer replacement courses.

The Issues API now includes descriptions for the status, triageState, and aggregateState values, and designates aggregateState as the recommended field to use.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Fastly's real-time analytics and historical stats API responses now include imgopto_input_image_bytes and imgopto_output_image_bytes, giving visibility into the input and output data volumes of Image Optimizer traffic.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

New --connect socks5:<host>:<port> (and socks4 variant) routes PolarProxy's outgoing connections through an upstream SOCKS4 or SOCKS5SOCKS5A proxy protocol that relays TCP and UDP traffic through an intermediary server without inspecting it, letting cyber tools tunnel connections through firewalls or pivot hosts with minimal overhead. proxy, enabling use cases such as decrypting and inspecting your own HTTPSHTTPSHTTP carried over an encrypted TLS connection. It protects traffic in transit and lets a browser verify the server's certificate. traffic before it is forwarded through Tor.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

$ PolarProxy --socks 1080 --connect socks5:127.0.0.1:9050 --nontls allow --pcapoverip 57012 --writeall

Adds --idletimeoutclient <seconds> and --idletimeoutserver <seconds> to close idle/'dead' connections after a configurable period, and --tlstimeout <seconds> (default 30s) as a dedicated TLS-handshake timeout separate from the existing TCPTCPA network protocol that delivers a reliable, ordered stream of bytes between two endpoints. It detects lost data and retransmits it.-handshake --timeout <seconds>.

$ PolarProxy --socks 1080 --idletimeoutclient 120 --idletimeoutserver 120 --tlstimeout 30

PolarProxy now accepts environment variables as an alternative to command-line arguments, using a PP_ prefix (e.g., PP_SOCKS maps to --socks), for cleaner config separation in systemd, Docker, Podman, Kubernetes and Helm deployments via a polarproxy.env file; the new --helpenv flag prints the full list of supported environment variables.

Ships with a docker-compose.yml for musl builds, or a systemd unit file with install-polarproxy-service.sh for non-musl Linux builds, to simplify initial deployment.

The --writeall flag extends PCAPPCAPA packet capture, recorded network packets that can be inspected later. Analysts use it to reconstruct communication and investigate protocol behavior or suspicious traffic.-over-IP output to include all proxied traffic, not just decrypted TLS sessions.

$ PolarProxy --socks 1080 --connect socks5:127.0.0.1:9050 --nontls allow --pcapoverip 57012 --writeall

The new go-openssl-fips container image for Go 1.27 selects its FIPS-validated cryptographic module at deployment time rather than at compile time, enabling a single build artifact to run across environments without recompilation.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Usage: FROM cgr.dev/chainguard/go-openssl-fips:latest.

$ FROM cgr.dev/chainguard/go-openssl-fips:latest

Automatically maps which AI agents exist across the SaaS stack, what they are connected to, and who is using them, covering platforms including Claude, Microsoft 365 Copilot, Salesforce Agentforce, and ServiceNow Now Assist.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

⤢ ⤢

Adds an OpenAI-compatible AI provider with support for all auth methods, letting teams point Tines at any OpenAI-compatible endpoint.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ This supports routing through Amazon Bedrock Mantle by configuring https://bedrock-mantle.<region>.api.aws/openai/v1 as a custom provider endpoint with API type OpenAI Responses and manually adding model IDs such as openai.gpt-5.6-luna.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

⤢ ⤢ ⤢ ⤢

Adds a New Connectors tab in Monitoring for connector observability, a Tines Units drill down, a Set currency option in the Savings tab within Monitoring, a New 'Fix failures' button and New 'Add tests' button on workflow-level monitoring, Jump through step execution failures navigation, and the ability to Set unit tests for workflow steps.

Adds a new MCP settings page for configuring 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. integrations, along with Support SSO route auth for MCP servers, MCP support for spaces using git sync, and an option for MCP server connectors to skip authentication.

Adds new OAuth 2.0 connector for ServiceNow, a OAuth 2.0 JWT bearer connector, support for JSON token request format in custom OAuth connectors, and connector status display for OAuth errors with tooltip surfacing. New service connectors include SendSafely connector, Greenhouse Harvest v3 OAuth connector, DatoCMS connector, Zscaler OneAPI OAuth connector, Jira Service Management connector, Gainsight connector with OAuth support, and Productiv connector. Also adds AWS RDS IAM authentication for database connectors and GitHub connector support for GitHub Enterprise Server.

Adds a Space level block self-approval setting for change control, a warning before pushing live when a link's access scope changed, the ability to rename new stories under change control without requiring approval, and change control comments enabling annotated approval workflows.

Adds a Search and filter option to audit logs, and a configurable audit log retention period so teams can control how long audit records are kept.

The Jobs API now supports performance_target set to STANDARD for one-time runs, alongside existing performance targets, to control cost vs.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ speed trade-offs at submission time.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Apache AirflowApache AirflowAn open-source workflow orchestration platform maintained by the Apache Software Foundation that lets developers schedule, monitor, and chain tasks as directed acyclic graphs, making it useful for automating data pipelines.'s DatabricksSubmitRunOperator also supports performance_target: STANDARD for one-time Databricks job submissions.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

$ curl -X POST https://<databricks-instance>/api/2.1/jobs/runs/submit \
  -H 'Authorization: Bearer <token>' \
  -H 'Content-Type: application/json' \
  -d '{"performance_target": "STANDARD", "run_name": "my-one-time-run", "tasks": [{"task_key": "main", "existing_cluster_id": "<cluster-id>", "notebook_task": {"notebook_path": "/path/to/notebook"}}]}'

Databricks-provided 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. connectors for Google Drive, Gmail, Google Calendar, Microsoft 365, Atlassian, Slack, and GitHub are now generally available, letting Genie One and Genie Code search and act on these external company tools while answering questions. Each user authenticates individually via OAuth so tokens are never shared across users, with access governed through Unity Catalog and Unity Gateway.

Databricks OpenSharing is now generally available for sharing objects federated via Lakehouse FederationLakehouse FederationA data architecture pattern that lets a single query engine read from multiple external data sources—databases, data lakes, warehouses—without moving or copying the data, reducing pipeline complexity for tools that need broad data access. without copying data: foreign Iceberg tables federated from external Iceberg catalogs can be shared with Delta Sharing recipients, including those using non-Databricks Iceberg clients, and foreign schemas and tables can now be read by recipients through OpenSharing.

Extends Elastic prebuilt rule coverage for Linux privilege escalationprivilege escalationGaining permissions beyond those an account or process initially had. It can turn limited access into administrator control or access to another user's resources. techniques mapped to 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. 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. T1548.001 (Abuse Elevation Control Mechanism: Setuid and Setgid), available in the detection-rules repository (SIEM) and the protections-artifacts repository (endpoint behavioral rules).⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Rules can be enabled in Kibana under Security -> Rules -> Detection rules, filtering by tag OS: Linux and tag Tactic: Privilege Escalation.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

⤢ ⤢

New rule inspects process.args (rather than process.name) for known setuid binary paths — /usr/bin/pkexec, /usr/lib/snapd/snap-confine, /usr/bin/newuidmap, and /usr/libexec/openssh/ssh-keysign — combined with process.args_count <= 2 to catch cases where the helper is passed as an argument to another privileged binary.

New rules key on the common escalation skeleton — an unprivileged process (uid != 0) executing from writable paths such as /tmp, /dev/shm, /var/tmp, /run/user/, or /home/ that becomes root — to cover page-cache and no-userland-tell kernel bugs like Copy Fail (CVE-2026-31431) and DirtyFrag (CVE-2026-43284) where no implementation-specific artifact exists.

⤢ ⤢

Adds guidance for syscall-level LPE visibility via Auditd Manager, requiring socket, splice, bind, and execve audit rules to surface the page-cache exploit family including Copy Fail and DirtyFrag.

New detection rule combining an euid-0 / non-root-real-user process shape, low process.args_count, and an interpreter or writable-path parent to catch abuse of su, sudo, pkexec, and passwd with near-zero false positives.

New rule uses an EQL stringcontains(process.executable, process.command_line) trick to catch abuse of uncommon or long-tail SUID binaries without maintaining a hard-coded name list.

Agent Classifiers evaluate AI agent risk at call time by writing rules over private-data sensitivity, untrusted-content exposure, and exfiltration capability, then routing matches to a hook, human approval, or block.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ An Observe mode lets admins log what a rule would catch before turning on enforcement.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

The Owner column, filter, and detail field on Findings are renamed to Assignee across the UI, API, webhooks, and audit log; consumers of the Finding object via API, webhook, or 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. integrations must update field and event-name references accordingly.

Admins can now set a custom bulk request entitlement limit from Governance > Forms, ranging from 1 to 100 with a default of 10.

Automation expressions that reference the bare workflow context, or pass it directly to a function, no longer compile — each expression must name the step and field directly. Already-running automations are unaffected until next edited and saved, at which point indirect-reference expressions must be rewritten to a static path.

System log events now include a groups[] field and correlation_uid, plus four new Identity & Access event classes: Account Change, Entity Management, User Access Management, and Group Management. Machine identities are now reported as the Service account type in these logs.

⤢

A new Resources tab for MCP servers lets admins review and approve static resources and URI templates discovered by C1; each approved resource gets its own read entitlement, governed separately from tool and toolset access.

A new Connector sync failing finding type triggers after two consecutive failed full syncs and auto-resolves on the next successful sync; it can be disabled per tenant.

Role mining gains a shared-access threshold setting, pre-action entitlement list fine-tuning, and optional routing of automatic profile enrollments through an access request task for approval.

⤢ ⤢

Clerk now supports the device_authorization_endpoint and grant type urn:ietf:params:oauth:grant-type:device_code, automatically advertised via 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. and OpenID ConnectOpenID ConnectOpenID Connect, a sign-in protocol built on OAuth 2.0. An application uses identity information from a trusted provider instead of maintaining a separate password for each user. metadata when the Device Authorization GrantDevice Authorization GrantAn OAuth 2.0 flow for devices that lack a browser or keyboard: the device gets a short code, displays it to the user, and polls the authorization server while the user approves on a separate device. is enabled.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ Clerk's Account Portal hosts user verification for device flows, displaying the OAuth application, requested scopes, an optional Organization selection, and approve/deny actions.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

⤢ ⤢

New Identity Management > Agents section (Public Preview) lets admins register and govern non-human AI agent identities with OAuth client credentials, grant group-based SSOSSOSingle sign-on, using one identity provider to sign in to multiple applications. It reduces separate passwords and centralizes account and access management. application access via Agent Groups, and manage their lifecycle independent of human user accounts.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

New Insights > Reports section adds a Custom Query Builder supporting manual report building and natural-language queries, filter operators (Equals, Contains, Is Null), and export in CSV and JSON formats with scheduling.

Access Risk Detection is now GA, providing behavior-based detection of unusual locations, impossible travel, unfamiliar devices/browsers, dormant accounts, and abnormal login volumes across the Admin Portal, User Portal, managed device logins, and SSO applications. It supports customizable risk factors, automatic 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.-based resolution, intelligent VPNVPNA virtual private network, an encrypted tunnel between devices or networks. It carries traffic across another network as though the endpoints had a private connection./gateway suppression, and email, Slack, and webhook notifications.

Device settings gain MCP Discovery and Token & Cost Tracking options under Enhanced Discovery Options, surfacing MCP servers and token/cost data from Cursor, Claude Code, OpenCode, Codex, VS Code, and Copilot CLI on managed Windows and macOS devices.

MCP servers previously created without an associated SSO application or bookmark now require one to be assigned in AI Gateway; access is no longer scoped by User Groups until an application is assigned. The custom app creation option has been removed from the Add 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. dialog — a + New App button now launches the standard Add New SSO Application wizard during server setup instead.

Admin Portal now includes Password Vault management: vault enablement, user/group enrollment, permission management (Connect, View, Edit, Manage), credential sharing, and password health monitoring via an Overview dashboard. Password Vault audit log entries were also added to Directory Insights.

Device Trust and conditional access policy support now extends to mobile devices managed by third-party MDMsMDMsMobile device management, centrally configuring and managing enrolled devices. Administrators use it to apply settings, distribute applications and enforce device security policies. (e.g., Jamf, Kandji, Google Workspace EMMEMMEnterprise Mobility Management: a category of software for enforcing policies on mobile devices, apps, and data, giving cyber tools a hook to monitor or control endpoints without managing the full OS.), registered via a shared secret with an SSO Extension profile in the existing MDMMDMMobile device management, centrally configuring and managing enrolled devices. Administrators use it to apply settings, distribute applications and enforce device security policies..

The JumpCloud Go browser extension gains Password Vault autofill, supporting launch-and-autofill for non-SSO websites and 2FA codes, credential secret viewing, and save-on-login/update/signup — with no device registration required.

Password Vault is now available in the User Portal left navigation, supporting credential and website management, sharing with Connect, View Secret, View Detail, and Manage permissions, and bulk import via CSV or Excel.

Adds three new cancel-request API endpoints — Update a request (Access Request V1), Update a request (Requests V2), and Update my request (My Requests) — in Beta, enabling programmatic cancellation of active and pending access requests from custom portals, ITSMITSMIT Service Management: a framework of practices for delivering and supporting IT services, covering ticketing, change control, and incident response. Cyber tools integrate with it to automate alert-to-ticket workflows. tools, CLIs, or chatbots.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

Adds a MALWARE_PROXIES IP service category to enhanced dynamic network zones, powered by Okta CyberDefense, covering proxy services linked to malware and botnet activity (including 911 S5, NSOCKS, iProxy, and BHProxies), now GA in Production. Also adds a new IP Service for enhanced dynamic network zones allowing zone policies to reference dynamically updated IP reputation feeds.

Enables admins to execute custom PowerShell scripts in on-premises Active DirectoryActive DirectoryMicrosoft's directory service for Windows networks that centralises authentication and authorisation; cyber tools target or integrate with it because it controls identity and access across most enterprise environments. environments via the Active Directory agent, invokable through Okta Workflows and the Okta public API Invoke a remote script on the AD agent; now GA in Preview.

Adds data.context.app.clientId and data.messageProfile.isResend fields to the telephony inline hook request payload, identifying the signing-in app and whether an OTPOTPOne-time password: a short-lived numeric or alphanumeric code valid for a single authentication attempt, generated via TOTP or HOTP standards, giving cyber tools phishing-resistant second-factor verification without stored shared secrets. is a resend within the same session.

{
  "data": {
    "context": {
      "app": {
        "clientId": "0oabc123def456"
      }
    },
    "messageProfile": {
      "isResend": true
    }
  }
}

Adds search by name or client ID, filter by type and status, a last-modified-date view, and CSV export to the Applications page, with programmatic access via the List all applications that match search filters API; available as Early Access.

Increases Access Request limits: users per task or question raised to 25 (from 10), entitlement bundles in an access level condition to 1,000 (from 100), groups in an access level condition to 1,000 (from 500), request type configuration lists per org to 250 (from 100), and request types per org to 750 (from 500).

Adds NFC authenticator support (self-service Early Access in Preview), enabling NFC-based hardware tokens as an authentication factor. Adds transports values to WebAuthnWebAuthnA web standard that lets a browser prove possession of a passkey or hardware key to a website. It is the browser-facing part of FIDO2 sign-in. authenticator enrollments (GA in Preview), exposing the physical transport (USB, NFC, BLEBLEBluetooth Low Energy, a wireless protocol maintained by the Bluetooth SIG for short-range communication at minimal power draw, making it useful for cyber tools that interact with IoT devices or proximity-based systems., internal) used during passkey registration. Supports prompting users to enroll a passkey during sign-in flows (GA in Preview), increasing passkey adoption without requiring admin-forced enrollment.