MCP Agent

A production-ready research agent using the Model Context Protocol (MCP), FastAPI, and Serper web search.

Features

• Async MCP server with web search tool (Serper API)

• FastAPI client for HTTP access

• Full test suite with pytest and Serper API mocking

Related MCP server: Deep Research Agent MCP Server

Setup

1. Clone the repo

2. Install dependencies:

   pip install -r requirements.txt

3. Set your Serper API key:

   • Create a .env file:

     SERPER_API_KEY=your_real_serper_api_key
     REQUEST_TIMEOUT=30
     MAX_RESULTS=20
     LOG_LEVEL=INFO

Usage

• Start FastAPI client (runs MCP server as subprocess):

  python -m client.fastapi_client

• POST /search for web search:

  { "query": "search term", "max_results": 10 }

• GET /health for health check

Testing

• Run all tests:

  pytest tests/test_e2e.py -v

• Run integration test (real Serper API):

  pytest -m integration tests/test_e2e.py

MIT License