PAGES
← all tools
◆ SAST / SCA / Code Security

deepsec

open-source

Deepsec is a security harness for finding vulnerabilities in your codebase powered by coding agents

Summary

deepsec is an open-source vulnerability scanner that uses AI agents to review large codebases for security issues that have sat undetected for a long time. It runs as a command-line tool: npx deepsec init walks you through picking a model and a payment method (your own API key or Vercel AI Gateway), then scans unattended, resuming automatically if interrupted, with --max-cost-usd and --max-duration flags to cap spend on runs that can otherwise reach thousands of dollars. It's aimed at application security teams and developers doing on-demand review of existing large-scale repos rather than continuous CI linting. It sits in the AI-driven code-review space, distinct from pattern-based static analyzers, fanning work out across worker machines for scale. First committed in 2026, it already has 9 contributors and 175 commits in the past year, suggesting active development.

Deepsec is a security harness for finding vulnerabilities in your codebase powered by coding agents

How deepsec helps

Which vulnerabilities in this codebase have been sitting undetected because no linter pattern catches them?

fans an AI review across every file in the repository at maximum reasoning depth, surfacing issues that pattern-based scanners are not built to see

Which parts of this application are actually reachable by an attacker versus just internal plumbing?

classifies each entry point by kind — endpoint, queue, cron, command line, webhook, agent tool — and by exposure, so findings can be weighed against how the code is actually reached

Did the last code review run get cut off before finishing, and does it need to start over?

tracks which files were already analyzed and resumes from there on the next run, so an interrupted scan never starts from the beginning

Is a finding from this scan the same one flagged last time, or something new?

assigns each finding a stable identity derived from its project, path, and title, so re-running the review recognizes repeat findings instead of duplicating them

Which of the flagged issues are worth a developer's time versus likely noise?

runs a revalidation pass that re-checks prior verdicts and reconciles them against the new scan, cutting the false-positive rate before anything reaches a report

How much is a review of this repository going to cost before it runs unattended for hours?

lets a run be capped by spend or by wall-clock time up front, since scans at maximum thinking levels can otherwise reach thousands of dollars on a large codebase

all 12 features, with the evidence for each →

Languages

Not documented

Features

12 capabilities across 5 areas

Built from deepsec's releases, docs, README and our own runs. “Seen” marks the month we first spotted a release the vendor didn't date. Group names and tags are our labels.

Compare with peers →

  • deepsec covers 1 of the 8 vulnerability assessment methods capabilities we track.

What deepsec offers you

  • Command-line interface2
Vulnerability Scanning 3 capabilities deepsec scans a codebase or running system for security weaknesses, covering a wide range of vulnerability classes and attack surfaces. It structures scanning around a formal inventory of exposed surfaces so coverage is deliberate rather than opportunistic.
Configurable agent reasoning depth shipped Lets the user control how deeply the agent reasons during a scan, trading speed against thoroughness. 1 release · 1 other source · first seen Aug 2026

release

  • Adds --thinking-level CLI flag to control agent reasoning depth during scans. commits-2026-07-09 · Aug 2026 · source · release history

example

  • Scan with cost and time limits npx deepsec init --max-cost-usd 100 --max-duration 2h
Attack-surface coverage analysis shipped Inventories all attack surfaces by kind and exposure level so the user can see what has and has not been scanned. 1 release · first seen Aug 2026

release

  • Introduces attack-surface coverage analysis (coverage.ts) with typed SurfaceKind (http, rpc, queue, cron, cli, webhook, agent-tool, other) and SurfaceExposure (public, authenticated, internal, mixed, unknown) inventories, enabling structured surface-aware scanning. commits-2026-07-11 · Aug 2026 · source · release history
SSRF detection shipped Detects server-side request forgery vulnerabilities across a broad range of SSRF patterns. 1 release · first seen Aug 2026

release

  • Broadens the SSRF matcher to cover a wider range of server-side request forgery patterns. commits-2026-07-09 · Aug 2026 · source · release history
Finding Management and Revalidation 3 capabilities deepsec tracks findings across repeated scan runs using stable identifiers and flexible verdict matching, so results can be compared over time without losing continuity. This lets teams track remediation progress and avoid treating the same issue as new on every run.
Permissive verdict reconciliation shipped Matches prior verdicts to current findings using several fallback strategies and records exactly how each match was made. 2 releases · first seen Aug 2026

release

  • Adds permissive verdict reconciliation for revalidate: verdicts now match by alias/findingId → exact title → normalized title → unique 1:1 remainder, with a matchedBy diagnostic recorded per match, replacing the previous exact-equality-only approach that silently dropped mismatched verdicts. commits-2026-07-11 · Aug 2026 · source · release history
  • Exports new processor primitives — reconcileVerdicts, buildAliasMap, expectedFindingsForBatch, resolveDuplicateRef, ensureFindingIds, attributionHeaders, setAttributionVersion, isGatewayBaseUrl — for use in custom integrations. commits-2026-07-11 · Aug 2026 · source · release history
Stable finding identifiers shipped Assigns each finding a stable, deterministic ID so the same issue is recognised as the same issue across multiple scan runs. 2 releases · first seen Aug 2026

release

  • Introduces stable, deterministic findingId values (SHA-256 of projectId + normalized path + title) assigned at append time and lazily backfilled on old records, so finding identity survives across revalidation runs. commits-2026-07-11 · Aug 2026 · source · release history
  • Introduces stable finding IDs and permissive verdict reconciliation in the revalidate workflow, backed by ensureFindingIds and reconcileVerdicts. commits-2026-07-09 · Aug 2026 · source · release history
Per-invocation prompt aliases for findings shipped Gives agents short, readable aliases for findings within a session so they can reference results without handling raw hashes. 1 release · first seen Aug 2026

release

  • Adds short per-invocation prompt aliases (F1…Fn) mapping to stable finding IDs so agents never see raw hashes in prompts. commits-2026-07-11 · Aug 2026 · source · release history
Workspace Setup and Initialisation 2 capabilities deepsec provides commands to prepare a project workspace for scanning without requiring manual configuration steps. Users can fully initialise and scan in one command or lay down scaffolding first and run automation separately.
One-shot workspace initialization shipped Sets up a new project workspace and starts scanning in a single command, with no manual configuration required. 2 releases · 1 other source · first seen Aug 2026

release

  • Ships one-shot Deepsec initialization (coordinator.ts) to scaffold a project from scratch without manual setup steps. commits-2026-07-11 · Aug 2026 · source · release history
  • Ships one-shot Deepsec initialization (PR #139) to bootstrap a workspace in a single command. commits-2026-07-09 · Aug 2026 · source · release history

example

  • Initialize and run first scan npx deepsec init
Scaffold-only workspace setup shipped Writes workspace configuration files without triggering a scan, so the user can review or edit them before running automation. 1 release · 1 other source · first seen Aug 2026

Also counts toward: Command-line interface

release

  • Documents --scaffold-only in docs/getting-started.md, including what it writes, what it skips, and how to resume the automated flow later, plus a Using local subscriptions section across models.md, FAQ, CLI help, and the scaffolded workspace README. commits-2026-08-09 · Aug 2026 · source · release history

example

  • Persist the local-subscription route in config so every subsequent preflight skips env-var checks automatically. ai: mode: local provider: local commits-2026-08-09 · Aug 2026 · source
Model Access and Authentication 2 capabilities deepsec supports multiple ways to authenticate with the AI models it uses for reasoning, from cloud API keys to local CLI sessions and brokered gateway access. This lets teams enforce consistent, auditable model traffic routing without changing how individual users log in.
Local subscription model authentication shipped Allows the user to rely on a machine-wide CLI login for model authentication instead of configuring cloud credentials separately. 1 release · 1 other source · first seen Aug 2026

Also counts toward: Command-line interface

release

  • Adds --model-auth local CLI flag (and interactive Use local subscriptions as option 4 in the model-access prompt) to configure a local model route, delegating model auth entirely to the machine-wide claude/codex login — no credential is resolved, injected, or verified. commits-2026-08-09 · Aug 2026 · source · release history

example

  • Bootstrap a workspace on a machine already logged in to the claude CLI, skipping all cloud credential setup. pnpm deepsec --model-auth local commits-2026-08-09 · Aug 2026 · source
AI gateway token brokering shipped Requires an AI gateway key for sandbox commands so all model traffic passes through a configured firewall gateway. 1 release · first seen Aug 2026

release

  • Sandbox commands under the local route still require AI_GATEWAY_API_KEY (an ambient key is accepted), preserving the firewall token-brokering requirement. commits-2026-08-09 · Aug 2026 · source · release history
Other 2 capabilities
AI-powered vulnerability scanning documented Uses AI models to review large codebases and identify hard-to-find security issues. 1 other source

Shared capability: Static application security testing

docs

  • Agent-powered vulnerability scanner that reviews large codebases for hard-to-find security issues using AI models.
Scan cost and time limits shipped Allows users to configure maximum cost and execution time limits for a scan invocation. 1 other source

example

  • Scan with cost and time limits npx deepsec init --max-cost-usd 100 --max-duration 2h

Lines in monospace are the tool's own words: help text parsed from its source, or an endpoint from its API document. Under release, a line is our summary of a dated release. Under docs, it is quoted from the tool's own documentation or website. Each links back to the source it came from.

Score card

  1. Shipping velocity releases a month, last 12 months
  2. Documentation quality release-notes rubric, 0–100
  3. Breaking-change discipline how rarely a release breaks users, and notice given
  4. Maintenance risk quiet time against its own cadence, maintainers
  5. Innovation & surface new capabilities, API and MCP surface
  6. Community health
  7. Contributor trend distinct committers, this year against last
  8. Maintainer concentration bus factor now against a year ago
  9. Issue backlog trend issues closed per issue opened, 12 months
  10. Responsiveness median time to a maintainer’s first reply

Nine scores for deepsec, each built from the releases, notes and API history this site tracks — and, on Researcher and Business, where it stands among every measured tool in its category and how that has moved.

Score cards are part of every paid plan.

See plans →

Each score is 0–100 on a published rubric, derived from data this site already shows. A dimension we cannot measure says so — it is never scored 0 and never enters a rank.

How this is scored
  • Shipping velocity — releases a month over the last 12 months, on a log scale where 16 a month scores 100. Needs 3 releases in the window.
  • Documentation quality — the release-notes rubric published on the analytics page, unchanged.
  • Breaking-change discipline — the share of releases carrying a breaking change (none scores 100, half or more scores 0). Only projects that label breaking changes are measured: a clean record cannot be told from an unlabelled one.
  • Maintenance risk — quiet time measured against the tool's own release cadence (quiet longer than 90 days and three typical gaps scores 0), plus the number of active maintainers and whether contributors are growing, where known. Higher is lower risk.
  • Innovation & surface — new capabilities in the feature ledger this year, API and MCP operations added, and the share of commits that are features: the mean of whichever we hold.
  • Community health (open-source tools only) — each is a direction, not a size: contributors this year against last, the bus factor now against a year ago, and issues closed per issue opened, where steady scores 50, doubled 100 and halved 0; plus the median time to a maintainer's first reply (an hour scores about 100, a month or more 0). A commercial card has no repository and carries none of the four.
  • Documented delivery (commercial tools) — when a vendor's documented new capabilities have slowed against its own baseline, that counts against the overall standing as heavily as two low dimensions. Steady delivery, or too little history to judge, leaves the standing alone; it is never a sixth score.
  • Open-source tools are scored on repository signals; commercial tools on their published changelog. Ranks compare tools within one category and one profile and need at least five measured peers; the overall standing needs three ranked dimensions. Best-in-breed or consolidator is a shape, not a quality, and is never scored.

Release-note detail over time

See how much detail deepsec puts in its release notes, quarter by quarter, and compare it with tracked tools in the same category. This history uses up to ten retained, reviewed notes per tool; sparse or blank older quarters may reflect that limit.

Checking release-note history…

Explore the full chart →

Releases

3 entries

  1. August 20263

    3 entries

    Project health

    Issue, contributor and maintainer activity in vercel-labs/deepsec, 2023-10 – 2026-09, as of 2026-09-25. Twelve-month figures first; the trend lines run the whole window.

    1. Contributors 10 +10 vs prior 12 months
    2. Maintainers 2 +2 joined · −0 left, last 12 months
    3. Bus factor 1 committers behind half the commits
    4. Community-filed 100% of issues opened, last 12 months
    5. Median time to close < 1 day first maintainer reply 5.9 hours
    6. Open backlog 23 of 35 issues ever filed
    1. Issues opened / month 0 · 0 communitymaintainers
    2. Open backlog 23
    3. Contributors / month 2
    4. Commits / month 3
    5. Stars 8,001
    Issue themes, last 12 months
    • bugs 37%
    • features 17%
    • other 46%

    Counts from the project's public repository and git history; no individuals are identified. Maintainers are the committers covering 80% of the last year's commits. how we measure this

    Add the badge to your README

    Maintain this project? The badge links readers straight to this page, so they can follow every release here.

    Tracked by Cyber Toolchain Choose a badge style and copy the code →
    browse all tools → score card for your toolchain →