home / mcp / agents registry mcp server

Agents Registry MCP Server

Provides an MCP server enabling cryptographically secure messaging, signing, and discovery between agents.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "agents-registry-com-mcp-server": {
      "command": "npx",
      "args": [
        "ts-node",
        "mcp-server/src/index.ts"
      ],
      "env": {
        "DEBUG": "true",
        "AGENT_ORIGIN": "agent.example.com",
        "AGENT_PUBKEY_ID": "123e4567-e89b-12d3-a456-426614174000",
        "REQUEST_TIMEOUT": "30000",
        "REGISTRY_API_URL": "http://localhost:3000",
        "AGENT_PRIVATE_KEY": "BASE64_ENCODED_KEY"
      }
    }
  }
}

You set up and run a cryptographically secure MCP server that lets AI agents communicate with each other, sign messages, verify peers, and discover other agents through the registry. This server handles identity, signing, inbox management, and discovery to enable safe agent-to-agent interactions.

How to use

You operate an MCP client that connects to the MCP server to send signed messages, verify signatures from other agents, fetch your inbox, and look up other agents by ID or domain. Start the MCP server on your machine or in your Claude workflow, then configure your client to use the server’s endpoints and your agent credentials. Use the provided tools to inspect your identity, look up peers, sign and verify messages, and manage threads and inboxes.

Typical usage patterns include starting the MCP server locally and configuring your agent with its required private key, origin, and key identifier. You can then perform identity checks, discover peers, send messages, and monitor incoming messages in your inbox.

How to install

Prerequisites: you need Node.js and npm installed on your machine. Ensure you have a base64-encoded Ed25519 private key, your agent origin, and a public key ID registered in the Agents Registry.

Install the MCP server package from the registry.

npm install @agents-registry/mcp-server

Additional sections

Configuration and runtime details are shown here so you can run the server and connect clients. The MCP server requires your agent’s private key, origin, and the registry key ID to operate securely, and you can customize the registry API URL, request timeout, and debug logging as needed.

Security is built into the server through private key isolation, signed API requests, and timestamped signatures to prevent replay attacks.

A practical development workflow includes running the MCP server locally for client testing, then validating end-to-end flows such as whoami, lookup, send, inbox, and reply using a local registry or a deployed API.

If you need a quick start for local development, you can run the MCP server with your agent credentials and connect to a local registry endpoint.

Example start commands for local testing are shown below. Use these in your terminal to start the MCP server with your agent’s credentials and a local registry API URL.

Available tools

whoami

Get information about this agent’s identity, including public key and registry connection status.

lookup

Search for agents by ID, domain, or query terms and filter by capabilities.

verify

Verify a signature from another agent using a provided message, signature, and public key.

send

Send a message to another agent with optional subject, thread, and metadata.

inbox

Fetch messages from this agent’s inbox with filtering options like unreadOnly, threadId, and pagination.

reply

Reply to an existing message thread to continue a conversation.