Everyone's a security professional.

Secfoo turns any AI coding agent into a disciplined security reviewer.

Same standards, same report, every run. Point Claude Code, Cursor, Codex, Copilot, Gemini CLI or another supported agent at a repo (or use a plain API key), pick your activities, and browse every result in one local dashboard.

secfoo run --skill security-architecture-review --agent claudeagentagygeminicodexcopilotopencodeapi▍
8 Security activities in the catalog
10 Agent adapters: 9 coding-agent CLIs plus a built-in API-key agent
SARIF 2.1.0 findings export for GitHub code scanning and IDEs
1 Local dashboard — no CDN calls while you read a report

How it works

01

Pick your activities

Choose one or more from the catalog — architecture review, threat modeling, SAST, SCA, secret scanning, and more. Run several together; they execute concurrently against the same target.

02

Point it at a target

A local directory, or a GitHub, GitLab (including self-hosted) or Bitbucket Cloud repo, plus optional Confluence links for extra context. Then choose who runs it: Claude Code, Cursor, Codex, Copilot, Gemini CLI and more, or an OpenAI, Anthropic or Gemini API key.

03

Browse the dashboard

Every assessment ever run, across every project, in one local web dashboard. Run secfoo serve and it's on your machine.

A 30-second look

The dashboard you get from secfoo serve

Program posture across every application, threat-modeling coverage and STRIDE, architecture design verdicts, and SAST findings that stay open until the code changes.

Recorded from a real local secfoo serve against public vulnerable-app targets.

Minutes, not weeks

A safety check, automated

Secfoo works alongside the AI tools your team already uses, and produces a consistent, standards-based security report on a piece of software in minutes rather than the days or weeks a manual review takes.

Every report follows the same structure — issues found, how serious each one is, and a clear pass or fail — so results can be compared, tracked and audited over time.

Built for CI

Gate the pipeline on findings and on spend

Fail a build on high-severity findings, cap what a scan may spend on AI, and get one JSON document back. Distinct exit codes tell CI whether the scan broke or the policy failed. Then upload the findings to GitHub code scanning as SARIF.

Set up a CI gate →
secfoo run --skill sast --skill secret-scanning \
  --agent api --target . \
  --fail-on high --max-cost 2.00 --json

# 0 clean · 1 run failed · 2 findings gate · 3 cost gate

secfoo findings export --format sarif \
  --project 1 -o findings.sarif

Eight activities, one orchestrator

See the full catalog →
security-architecture-review

Security Architecture Review

Controls matrix, design-principles verdict, CSA CCM v4 conformance, and a Design verdict for milestone gating.

threat-modeling

Threat Modeling

STRIDE (+LINDDUN), attack chains, every threat dispositioned, and an explicit residual risk statement.

sast

SAST — Static Code Analysis

Code-level findings with taint paths, CWE/OWASP mapping, a CVSS v3.1 score computed from the vector, and illustrative fix diffs.

sca-reachability

SCA — Reachability & Upgrade Triage

Dependency inventory, a reachability verdict per risk checked against real OSV.dev advisories, and a grouped upgrade plan.

secret-scanning

Secret Scanning

Exposed credentials across source, config, git history, and linked Confluence pages.

deployment-readiness

Deployment Readiness

Production security and operational readiness review before you ship.

Why teams choose Secfoo

Agent-agnostic

Claude Code, Cursor, Codex, Copilot, Gemini CLI, OpenCode, Antigravity, Droid or Devin. Secfoo drives whichever is already on your PATH, with an optional --model. No agent? Use an API key.

Local by default

Reports, the dashboard, and diagram rendering all run on your machine. No account to create before your first scan.

Honest about approximations

Program dashboards are built from what a report can actually state. Where the data can't support a claim, Secfoo shows zeros or hatching. A run with no reported cost is "unpriced", never $0.

Findings that follow the code

SAST, SCA and secret findings are keyed by the code they point at, not the agent's wording, so a rescan closes what you fixed and keeps the rest open.

Open source

MIT, no vendor lock-in. Read the skill definitions, the report contracts, and the dashboard code directly on GitHub.

Everyone's a security professional

Expertise built into the product, not required of the person running it

A developer with no security training gets the same quality of review a specialist would produce. Security stops being a bottleneck waiting on a small expert team and becomes something every engineer does as part of their normal work.

Built for your role

Security architects

Gate milestones with evidence

Design verdicts, CSA CCM v4 domain conformance, and a gate-decision mix (approve / with conditions / reject) you can point to at a review.

AppSec & threat modelers

Track disposition, not just findings

STRIDE coverage per asset class, an explicit residual-risk statement, and recorded risk acceptances that survive the next re-run.

Platform & DevOps

One CLI, wired into the pipeline

Severity and cost gates with CI-friendly exit codes, --json output, SARIF for code scanning, and secfoo cost to see AI spend by agent, skill or project.

For larger organisations

One view across every team

A central portal gathers every team's results into a single view, so management can see risk and coverage across the whole business without chasing individual reports.

See Cloud & Enterprise editions →

Install in under a minute

pip, npm, a universal shell script, or Docker — pick whichever fits how you work. They're all the same tool underneath.

See install options