Agent Services MCP server

agent-services-mcp

A single thin MCP (Model Context Protocol) server that exposes three existing services as discoverable tools, so AI agents and MCP-compatible clients can find and use them through one connection:

returns an Ed25519-signed receipt.

• provenance-receipts — certifies content origin;

published rubric; returns an Ed25519-signed score receipt.

• quality-gate — scores content quality against a

agent took (agent accountability); returns an Ed25519-signed audit receipt.

• agent-action-audit — signs a tamper-evident audit receipt for an action an

It is a thin wrapper. Every tool forwards an HTTP call to the underlying

Worker and returns its response verbatim. It does not reimplement signing,

scoring, or payment logic — those live in the underlying services. The honesty

about what each service proves carries through to the tool descriptions.

Quickstart — your first (free) call in ~2 minutes

npm install && npm run build
node examples/free-call.mjs       # connects to the LIVE services and calls a free tool

examples/free-call.mjs runs an MCP client against this server (pointed at the live deployments) and calls getqualityrubric and verifyaudit — both free, no wallet needed. To wire the server into an MCP client (Claude Desktop / Claude Code style), see Connecting an MCP client** below.

Free vs paid at a glance: verifyprovenance, verifyquality, verifyaudit, and getqualityrubric are free. certifyprovenance, scorequality, and auditaction are paid (an x402 USDC micropayment on Base) — see Calling paid tools for the two-step payment flow and a working example.

What the wrapped services prove (and do not)

was issued by the service's key. The generatormetadata is caller-attested — it proves you claimed* it, not that a specific model ran. Not AI-detection, not a truth guarantee.

• Provenance: proves the content is unmodified (SHA-256 hash) and the receipt

completeness, internal consistency, obvious-error freedom). Not absolute truth, not an external standard, not a fact-check. Read the rubric via the getqualityrubric tool.

• Quality: a reproducible score against the published rubric (clarity,

and unaltered since issue (tamper-evident). The action, actormetadata, and context are caller-attested — it proves you claimed this record, not that the agent's claim is true. An accountability/audit tool, not** a lie-detector.

• Audit: proves the action record is genuine (issued by the service's key)

Tools

Shortened. The full README is on GitHub.