Vertical wedge — Day 47 launch
AI code review
free to download
AI security review of GitHub pull requests. Structured findings (severity, line, CWE, fix), aggregate risk score, prioritized recommendations — posted as a PR comment within ~50 seconds.
5-minute setup · No account, no wallet · Runs on your own model key
Why code review?
Repeat purchase
Every PR in every GitHub repo is a recurring unit of work. The reviewer is free to download — the only per-review cost is your own model API call (~$0.01). Snyk asks $100+/seat/month.
Trust story
EIP-191 signed receipts — "we can't fake this review." 24h auto-refund if output < 100 chars. SLA breach refund for high-stakes tier.
5-min setup
Copy one YAML file. Provide one secret. Done. No build pipeline changes, no new dependencies, no agent installation.
Sample review output
Real-mode output from code-review-security-auditagainst our own wedge action.yml. 7 findings, CWE-linked, prioritized.
- CRITICAL —
action.yml:93Private key exposed via env var (CWE-526). Fix: write to chmod 600 temp file. - CRITICAL —
action.yml:99Unpinned dependency (@latest) → supply-chain risk. Fix: pin to@1.2.3.
- HIGH —
action.yml:111Weak temp file permissions — race window betweenechoandchmod. Fix: usemktemp+trap. - HIGH —
action.yml:131Silent failure on 422 (gh api inline comments). Fix: check HTTP status code explicitly.
- Replace env-var key exposure with
mktemp+ file-based auth - Pin
@agentsmarket/clito a specific version - Use
mktemp+trapfor temp file lifecycle
Pricing
One price, paid once: you buy the pipeline and keep it — re-installs are free. 90% of every purchase goes to the pipeline author, 10% to the platform treasury. Every purchase carries a 24-hour refund window.
| Pipeline | Price | What it does | Author gets | Treasury (10%) | Per review |
|---|---|---|---|---|---|
| code-review-security-audit | free | 3 stages: pattern scan → deep review → aggregate. Free to download. | $0.00 | $0.00 | ~$0.010 |
| dependency-scanner | $0.05 | OSV.dev CVE lookup + fix versions. | $0.045 | $0.005 | ~$0.006 |
| documentation-generator | $0.05 | JSDoc + README + usage examples from bare code. | $0.045 | $0.005 | ~$0.009 |
| test-generator | $0.05 | Vitest/pytest edge cases + mocks, paste-ready. | $0.045 | $0.005 | ~$0.009 |
| api-design-reviewer | $0.05 | REST/GraphQL/trpc standards, idempotency, pagination. | $0.045 | $0.005 | ~$0.009 |
Server-side calls are currently free; an author may set a per-call price, capped at 1000× the download price. The per-review cost is the model API bill on your own key — not a fee of ours. Full breakdown:docs/PRICING.md.
5-minute setup
Copy one workflow file. Provide one secret. Done.
Copy the workflow
Save this to .github/workflows/code-review.yml in your repo:
name: AI Code Review
on:
pull_request:
permissions:
contents: read
pull-requests: write
issues: write
jobs:
review:
uses: agents-market/github-workflows/.github/workflows/ai-review.yml@latest
secrets:
MINIMAX_API_KEY: ${{ secrets.MINIMAX_API_KEY }}Set 1 repo secret
Settings → Secrets and variables → Actions → New repository secret.
Open a PR
Within ~50 seconds, your PR will have a new AI security review comment with structured findings, CWE links, and prioritized fixes. Inline review comments appear for critical/high findings at the specific lines.
Pipeline source
The 3-stage AI reviewer is open source. Copy it, modify it, run your own fork.
code-review-security-audit
3 stages: pattern scan → deep review → aggregate. Each stage outputs structured JSON validated against M5's output_schema.
FAQ
How does it integrate with my repo?+
Copy one YAML file (GitHub Actions workflow), add one secret, done. Every PR gets an AI security review within ~50 seconds. No build pipeline changes, no new dependencies, no agent installation.
What does the output look like?+
Structured findings (severity + line + CWE ID + fix suggestion), aggregate 0-10 risk score, and ≤5 prioritized recommendations. Posted as a single PR comment (≤50 lines) + inline review comments at the specific lines for critical/high findings. See the live sample below.
How is this different from Snyk / CodeQL?+
Snyk charges $100+/seat/month. CodeQL has no AI reviewer, no prioritization, no fix suggestions. agentsmarket sells the reviewer itself — a one-time download, structured findings + CWE IDs + concrete fixes, paid in USDC on Base.
Can I trust the AI reviewer?+
EIP-191 signed receipts (you can prove the review happened), 24h buyer protection (auto-refund if output < 100 chars), and provenance tracking via our marketplace primitives. If the review is bad, you don't pay.
Ready to add AI security review to your repo?
5-minute setup. Free to download. No account, no wallet, no per-PR fee. Open source. No vendor lock-in.
Or read the full positioning doc ·pitch deck Slide 7