npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@fractalai/agent-passport-mcp

v0.2.3

Published

Model Context Protocol (MCP) server to ISSUE and independently VERIFY verifiable agent passports: ML-DSA-65 (NIST FIPS 204) signature verification + FRC-55 on-chain agent lookup on the FractalAI L1. Works with LangChain, CrewAI, AutoGen, OpenAI Agents — a

Readme

Agent Passport — MCP server

The free tool a developer drops in, in ~5 lines, to issue and independently verify a verifiable passport for an AI agent — over the Model Context Protocol, so it works in LangChain, CrewAI, AutoGen, OpenAI Agents, Claude Desktop — anything that speaks MCP.

It talks to the real, public FractalAI deployment: the live L1 JSON-RPC node at https://api.fractalai.net.co (FRC-55 agent registry) and the web endpoints at https://fractalai.net.co (attestations, receipt key, x402). Nothing here is mocked.

npx @fractalai/agent-passport-mcp

Tools

| Tool | Cost | What it REALLY does today | |------|------|---------------------------| | verify_passport | free | Given an attestation permalink / 0x hash (and optionally the receipt JSON): recomputes the content hashes, reconstructs the exact canonical signed message, verifies the ML-DSA-65 (NIST FIPS 204) signature 100% offline against the published /api/x402/receipt-key, verifies the Dilithium-2 signature server-assisted with a tamper negative-control, and reads the agent's FRC-55 on-chain state. Returns { valid, algorithm, checks[], anchored, decision, agent }. | | issue_passport | x402 ($0.05 USDC) | Binds an agent_id + a signed assertion + a re-verifiable permalink: resolves the agent on the FRC-55 registry (if it's an on-chain id) and seals the assertion through the real x402 attest-decision flow (Dilithium-2 signed, Black-Box chained, anchored on the L1). Without a payment_tx_hash it returns the payment challenge — it never fabricates a passport. | | lookup_agent | free | Public FRC-55 registry read on the live node: fractal_getAgent for one agent (name, owner, status, model, total_calls, rating, total_ratings, created_at) or fractal_listAgents for all, always with fractal_getAgentStats. | | list_fractalai_proofs | free | Self-describing discovery: reads the canonical /.well-known/x402.json live and returns the full proof catalog (price, algorithm, endpoint, what each proves) — one call maps the whole x402 surface and auto-heals if it changes, since it reflects the live source of truth rather than a hardcoded copy. |

x402-paid tools (v0.2.0) — post-quantum proofs, pay-per-call

The 7 FractalAI x402 routes are surfaced as auto-payable MCP tools. An x402-aware wallet client signs an EIP-3009 authorization and pays in USDC on Base using the standard _meta["x402/payment"] convention. Each served signature is ML-DSA-65 / Dilithium-3 (NIST FIPS 204) and verifiable offline; each also attempts an on-chain anchor on FractalAI's L1 (the anchor signature is Dilithium-2 / ML-DSA-44-class, pending the gated consensus cut-over — see "Honest scope"). Uses the FIPS-204-standardized algorithm via audited @noble/post-quantum (not a FIPS-validated CMVP module). Call any tool without payment first to get the exact price + payment challenge.

| Tool | Price | What it does | |------|------|---------------| | seal_ml_dsa | $0.02 | Post-quantum signature over any message/hash — durable, quantum-safe, offline-verifiable. | | verify_proof | $0.01 | Verify a signature (auto-detects ML-DSA-65 or Dilithium-2). The floor primitive for any proof pipeline. | | verify_agent_kya | $0.03 | Know-Your-Agent: verify another agent's credential + live FRC-55 reputation before you trust/pay it. | | notarize_decision | $0.05 | Notarize an AI decision (input, output, model) → auditable, non-repudiable, on-chain (EU AI Act Art. 12). | | prove_provenance | $0.05 | Provenance seal over a dataset/content hash + drand timestamp + on-chain anchor. | | attest_training | $0.05 | C2PA training-provenance manifest for a model, post-quantum signed. | | seal_cbom | $0.05 | Seal a CycloneDX CBOM (crypto inventory) — CNSA 2.0 / procurement PQC-readiness disclosure. |

Payment note: paid tools accept the standard _meta["x402/payment"] (auto-pay) and also a plain payment string arg (base64 X-PAYMENT) as a universal fallback. Nothing is fabricated — the tools proxy to the live routes and relay exactly what they return, honest scope included.

Honest scope (read this — project Regla 4)

A valid passport proves an identity + a specific attested assertion are cryptographically intact over the exact signed bytes, that the record is internally consistent, and that it is anchored on FractalAI's L1. It does NOT prove:

  • that the assertion is true, or that the agent's decision/action was correct;
  • any decentralized / consensus-scored reputationissue_passport does not register reputation;
  • lookup_agent ratings are the registry's own counters as reported by the node, not an independent audit.

Signature checks are labeled by trust model: [independent] = computed locally, does not trust the FractalAI server (the offline ML-DSA-65 check, content hashes); [server-assisted] = calls a FractalAI endpoint (the Dilithium-2 /api/proofmind/verify check). The ML-DSA-65 verification uses the FIPS-204-standardized algorithm via the audited @noble/post-quantum library — this is not a FIPS-validated (CMVP) module.

What is real today vs roadmap

  • Real today: offline ML-DSA-65 verification against the published key; server-assisted Dilithium-2 verification + tamper control; FRC-55 on-chain agent reads; the real x402 402→pay→seal flow for issuing.
  • Roadmap (not claimed as done): server-issued attestations are currently self-signed by the server key and anchored on FractalAI's own L1; full on-chain consensus validation and a decentralized, consensus-scored reputation are future milestones. The ml_dsa65 detached signature is delivered to pilot integrations today (pass it via the receipt argument) and is being rolled into the public API.

Use it

You don't need to install or build anything — npx fetches and runs the published package directly (see the config block above). No account, no API key.

Contributing (building from source)

Only needed if you're working on the server itself, inside the FractalAI monorepo:

cd integrations/agent-passport-mcp
npm install
npm run build      # tsc → dist/
npm test           # handshake: server stands up, tools respond, ML-DSA-65 round-trips

Configure in an MCP client

Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "agent-passport": { "command": "npx", "args": ["-y", "@fractalai/agent-passport-mcp"] }
  }
}

Environment

  • FRACTALAI_BASE_URL — web endpoint base (default https://fractalai.net.co).
  • FRACTALAI_RPC_URL — L1 JSON-RPC node (default https://api.fractalai.net.co).
  • AGENT_PASSPORT_AUTODEMO — on connect, the server makes one real, free call to the live node (fractal_getAgentStats) and logs the result to stderr, so you can see immediately that this isn't mocked. Fire-and-forget — it never delays or blocks the MCP handshake. Set to 0 to disable.

5-line examples (per framework)

All of these run the same MCP server (npx @fractalai/agent-passport-mcp) over stdio and call the tools.

LangChain (Python)langchain-mcp-adapters:

from langchain_mcp_adapters.client import MultiServerMCPClient
client = MultiServerMCPClient({"agent_passport": {"command": "npx", "args": ["-y", "@fractalai/agent-passport-mcp"], "transport": "stdio"}})
tools = await client.get_tools()   # verify_passport, issue_passport, lookup_agent
agent = create_react_agent(model, tools)
await agent.ainvoke({"messages": "Verify passport https://fractalai.net.co/attestation/0x…"})

CrewAI (Python)crewai-tools MCP adapter:

from crewai_tools import MCPServerAdapter
from mcp import StdioServerParameters
params = StdioServerParameters(command="npx", args=["-y", "@fractalai/agent-passport-mcp"])
with MCPServerAdapter(params) as tools:
    Agent(role="Auditor", tools=tools, goal="Verify agent passports", backstory="...")

AutoGen (Python)autogen-ext MCP workbench:

from autogen_ext.tools.mcp import StdioServerParams, mcp_server_tools
params = StdioServerParams(command="npx", args=["-y", "@fractalai/agent-passport-mcp"])
tools = await mcp_server_tools(params)   # verify_passport / issue_passport / lookup_agent
agent = AssistantAgent("auditor", model_client=client, tools=tools)

OpenAI Agents SDK (Python):

from agents import Agent
from agents.mcp import MCPServerStdio
async with MCPServerStdio(params={"command": "npx", "args": ["-y", "@fractalai/agent-passport-mcp"]}) as server:
    agent = Agent(name="Auditor", instructions="Verify and issue agent passports.", mcp_servers=[server])

Independent, no-framework verification

Because a receipt is designed for third-party verification, you don't even need this server (or any FractalAI code) to check one — just @noble/post-quantum and the public key directory:

import { ml_dsa65 } from '@noble/post-quantum/ml-dsa.js';

// 1. Pin the trusted key(s) from the signed, ML-DSA-65-governed directory (fetch once, cache):
const dir = await (await fetch('https://fractalai.net.co/.well-known/x402-receipt-keys')).json();
// (dir.signature / dir.directory_public_key let you verify the directory itself is unmodified —
// see @fractalai/pqc-agent-receipts-conformance for a full conformance-checked example.)

// 2. Given ANY served receipt — e.g. the response of a paid /api/x402/* call — verify offline:
const { signature, signed_message, public_key } = receipt; // fields returned by every x402 route
const valid = ml_dsa65.verify(
  Buffer.from(signature, 'base64'),
  new TextEncoder().encode(signed_message),
  Buffer.from(public_key, 'base64'),
);
console.log('receipt valid:', valid); // true only if the exact bytes were signed by that key

For a battle-tested, adversarially-vectored version of this same check (genuine + tampered + wrong-key forgery vectors across x402/SAR/ACP/JOSE/W3C-VC profiles), see @fractalai/pqc-agent-receipts-conformance (npx pqc-receipt-check).

License

MIT