Custom Agent Integration with MCP Server

This project demonstrates how to set up an MCP (Model Context Protocol) server that acts as a bridge between Claude (or any MCP-compatible client) and the Synthasaurus API, allowing you to query custom AI agent (e.g., Co:Driver) and return responses.

How it Works

• The server is built using FastMCP from the mcp.server.fastmcp package.
• The api_call tool is exposed to the MCP client (e.g., Claude). When called, it:
  1. Starts a new conversation with the Synthasaurus API.
  2. Sends the user's message to agent (e.g., Co:Driver) via the API.
  3. Polls for the AI agent's response if the API responds asynchronously.
  4. Returns the agent's response back to the MCP client.

Environment Variables

Set the following environment variables before running the server:

• AUTH_TOKEN: Your API authorization token.

Usage

1. cd mcp-server-demo
2. Install dependencies (using uv):

   uv add "mcp[cli]"

3. Running the standalone MCP development tools

    uv run mcp
   

4. Set the required environment variables (see above).
5. To add tool in claude app:

   mcp install main.py

6. you can test it with the MCP Inspector:

     mcp dev server.py
   

7. Connect to the server using an MCP-compatible client (e.g., Claude) and use the api_call tool to send a message to your agent.

Example

Send a message to your agent via Claude or another MCP client:

• Tool: api_call
• Input: "What is co:driver?"
• The server will return the response from your agent via the Synthasaurus API.

---

See mcp-server-demo/main.py for implementation details.