vouch-mcp 0.1.0

Model Context Protocol server that issues Vouch Credentials to authorize AI agent tool calls.

vouch-mcp

A Model Context Protocol (MCP) server that issues Vouch Credentials so an AI agent can cryptographically authorize the actions it takes.

MCP standardized how agents call tools. It does not say who is calling, or on whose authority. vouch-mcp adds that layer: every signed action carries a W3C Verifiable Credential with an eddsa-jcs-2022 Data Integrity proof, and optionally a delegation chain back to an accountable human principal.

Install

pip install vouch-mcp

This pulls in vouch-protocol[mcp], including the official MCP SDK.

Configure

The server reads two environment variables:

• VOUCH_PRIVATE_KEY the agent's private key (JWK JSON string).
• VOUCH_DID the agent's DID, e.g. did:web:agent.example.com.

Generate a development identity:

from vouch import generate_identity
kp = generate_identity()
print(kp.private_key_jwk)   # set as VOUCH_PRIVATE_KEY

Run

VOUCH_PRIVATE_KEY='...' VOUCH_DID='did:web:agent.example.com' vouch-mcp

Register it with an MCP client (Claude Desktop, Cursor, ...) as an stdio server running the vouch-mcp command.

Tools

Tool	What it does
sign_action(action, target, resource)	Issue a credential authorizing one action. Returns compact JSON to attach as a Vouch-Credential header.
create_session(purpose, valid_seconds)	Issue a longer-lived credential covering multiple actions.
get_identity()	Return the agent's DID.

Verify on the receiving side

from vouch import Verifier
ok, passport = Verifier.verify_credential(received_json, public_key=agent_pubkey)

License

Apache-2.0.