Proof Of Commitment MCP Server

Proof of Commitment

Stars lie. Behavioral signals don't.

An MCP server and web tool that scores npm packages, PyPI packages, and GitHub repos on behavioral commitment — signals that are harder to fake than stars, READMEs, or download counts.

The supply chain problem

Four packages in a typical Node.js project are CRITICAL right now:

• chalk — 412M downloads/week, 1 maintainer
• zod — 160M downloads/week, 1 maintainer
• lodash — 147M downloads/week, 1 maintainer
• axios — 100M downloads/week, 1 maintainer (attacked April 1st, 2026)

Stars and README quality don't surface this. Behavioral signals do.

Try it now

Terminal (zero install):

npx proof-of-commitment axios zod chalk
# or scan your own project:
npx proof-of-commitment --file package.json
# NEW: scan ALL transitive dependencies via lock file:
npx proof-of-commitment --file package-lock.json   # npm
npx proof-of-commitment --file yarn.lock           # yarn
npx proof-of-commitment --file pnpm-lock.yaml      # pnpm
# PyPI too:
npx proof-of-commitment --pypi litellm langchain requests

Web demo (no install): getcommit.dev/audit — paste your packages, see risk scores in seconds.

MCP server (zero install):

{
  "mcpServers": {
    "proof-of-commitment": {
      "type": "streamable-http",
      "url": "https://poc-backend.amdal-dev.workers.dev/mcp"
    }
  }
}

Add to Claude Desktop, Cursor, Windsurf, or any MCP-compatible AI tool. Then ask:

"Audit my package.json for supply chain risk" "Score axios, zod, chalk, lodash — which is highest risk?" "Is vercel/ai actively maintained?"

GitHub Action

Add supply chain auditing to any CI pipeline — auto-detects packages from package.json or requirements.txt, posts results as a PR comment, writes to GitHub Step Summary, and optionally fails on CRITICAL packages.

# .github/workflows/supply-chain-audit.yml
name: Supply Chain Audit
on: [push, pull_request]

jobs:
  audit:
    runs-on: ubuntu-latest
    permissions:
      pull-requests: write   # needed for PR comments
    steps:
      - uses: actions/checkout@v4
      - uses: piiiico/proof-of-commitment@main
        with:
          fail-on-critical: false   # set true to block merges
          comment-on-pr: true       # posts audit table directly on the PR

When comment-on-pr: true (default), the action automatically posts the audit table as a comment on the pull request — and updates the same comment on re-run, so you don't get comment spam. Reviewers see the risk table without leaving the PR.

Inputs:

Outputs: has-critical, critical-count, audit-summary (markdown table, also written to Step Summary).

Free vs Pro: Without an API key, packages are audited one at a time (with delays to respect rate limits). With a Pro API key, all packages are audited in a single batch request — faster and with higher monthly limits.

Example PR comment / Step Summary output:

| Package | Risk        | Score | Maintainers | Downloads/wk | Age   |
|---------|-------------|-------|-------------|--------------|-------|
| chalk   | 🔴 CRITICAL | 75    | 1           | 380M         | 12.7y |
| zod     | 🔴 CRITICAL | 83    | 1           | 133M         | 6.1y  |
| axios   | 🔴 CRITICAL | 89    | 1           | 93M          | 11.6y |

README Badges

Add a commitment score badge to any package you maintain or depend on:

![commit score](https://poc-backend.amdal-dev.workers.dev/api/badge/npm/YOUR-PACKAGE)

Examples:

Colors: 🟢 healthy (75+) · 🟡 good (60–74) · 🟡 moderate (40–59) · 🟠 high risk (<40) · 🔴 CRITICAL (single maintainer + >10M downloads/week)

Badges are cached 5 minutes at Cloudflare's edge. No API key needed.

REST API

No API key. No install.

curl https://poc-backend.amdal-dev.workers.dev/api/audit \
  -X POST \
  -H "Content-Type: application/json" \
  -d '{"packages": ["axios", "zod", "chalk", "lodash", "express"]}'

{
  "count": 5,
  "results": [
    {
      "name": "chalk",
      "ecosystem": "npm",
      "score": 75,
      "maintainers": 1,
      "weeklyDownloads": 398397580,
      "ageYears": 12.7,
      "trend": "stable",
      "riskFlags": ["CRITICAL"]
    },
    ...
  ]
}

7 MCP tools

What the score measures

Each package is scored 0–100 across:

• Longevity — How long has the package existed? Abandoned packages get reactivated for attacks.
• Maintainer depth — Single maintainer + millions of weekly downloads = the attack surface LiteLLM exploited.
• Release consistency — Regular releases signal active oversight. Long gaps = vulnerability accumulation.
• Download trend — Growing packages attract more scrutiny (and attacks). Stable = lower profile.

Risk flags:

• CRITICAL — single maintainer + >10M weekly downloads (exact LiteLLM/axios attack profile)
• HIGH — package <1yr old + rapid adoption
• WARN — no release in 12+ months

Real data points

chalk     — score 75, 1 maintainer, 411M/week  ⚑ CRITICAL
zod       — score 83, 1 maintainer, 159M/week  ⚑ CRITICAL
lodash    — score 87, 1 maintainer, 146M/week  ⚑ CRITICAL
axios     — score 86, 1 maintainer, 100M/week  ⚑ CRITICAL (attacked Apr 1 2026)
express   — score 97, 5 maintainers, 92M/week
litellm   — score 74, 1 maintainer           ⚑ CRITICAL (supply chain attack Mar 2026)

Why behavioral signals

The LiteLLM attack (March 2026) and axios attack (April 2026) followed the same pattern: stolen credentials → malicious package pushed → 97M+ machines exposed. Both packages scored CRITICAL by these metrics before the attacks.

Declarative signals (stars, README quality, CI badges) don't capture this risk. Behavioral commitment does.

Stack

Roadmap

Planned, not promised. The project is early-stage — contributions welcome on any of these.

See open issues for things you can help with today.

The broader vision

Supply chain auditing is the first tool. The underlying primitive is a commitment graph — behavioral signals that replace content-based trust across any domain.

When content is free to fake (reviews, stars, READMEs), commitment becomes the signal. A maintainer who has shipped 847 releases over 12 years is a different kind of commitment than one who published once in 2023.

The same logic applies to websites, businesses, and AI agents. Two card networks have independently named this gap: Mastercard Verifiable Intent §9.2 explicitly lists behavioral trust as "not covered." Visa TAP identifies agents without answering whether to trust them.

Proof of Commitment is the trust layer they're pointing at.

→ getcommit.dev

Run locally

bun install
bun run dev:backend     # local server with SQLite
bun run test:e2e        # E2E test with mock World ID

Deploy:

bun run deploy          # deploys to Cloudflare Workers