Skip to content

Use PeerX agents over MCP

The Model Context Protocol (MCP) is an open standard for letting an AI assistant call outside tools. PeerX serves each agent your organization can use as an MCP server, so a client such as Claude Desktop, Cursor or VS Code can call it.
Create a developer account

Set it up

  1. Create a developer account and a key

    Your client authenticates with a token from your developer account.

  2. Find the server you want

    The MCP documentation in your account lists the servers your organization can use and the tools each one has.

  3. Try its tools in the playground

    Connect, call a tool by hand and read what it returns before any client depends on it.

  4. Add the server to your client

    Put the server address and your token in the MCP settings of Claude Desktop, Cursor, VS Code or your own client.

How a call works

A client sends JSON-RPC 2.0 messages over HTTPS to the server’s address, with your token as a bearer token. A server only answers tokens from its own organization. The request below asks a server which tools it has.

Example request: list a server’s tools
curl -X POST https://api.peerx.ai/api/mcp/YOUR_SERVER_ID \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{ "jsonrpc": "2.0", "id": 1, "method": "tools/list" }'

Each server’s tools come from its agent’s definition, so they differ from agent to agent. For plain HTTP calls without an MCP client, use the REST API.