MCP Server & LLM Integration

A practical guide to building MCP servers with FastMCP and integrating them with any LLM using mcp-use.

Project Structure

• server/ - Basic weather MCP server using STDIO transport
• mcpserver/ - Weather MCP server with SSE/HTTP transport, includes client examples and Streamlit UI
• mcpagent/ - Simple MCP servers (math and weather) for agent integration examples
• graph_agent/ - LangGraph integration examples
  • main.ipynb - Weather tool agent with LangSmith monitoring
  • main.py - Competitor ads flow analysis using Playwright and DuckDuckGo search MCP servers
• app.py - Browser automation example using mcp-use with Groq LLM

Quick Start

# Install UV (if needed)
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install dependencies
uv sync

# Create .env file
echo "GROQ_API_KEY=your_key_here" > .env
echo "GEMINI_API_KEY=your_key_here" >> .env

# Run the weather MCP server
uv run python server/weather.py

# Run the LLM agent (in another terminal)
uv run python app.py

Resources

• MCP Spec: https://spec.modelcontextprotocol.io/
• FastMCP: https://github.com/jlowin/fastmcp
• mcp-use: https://github.com/ergut/mcp-use
• LangChain: https://python.langchain.com/
• LangGraph: https://docs.langchain.com/oss/python/langgraph/overview
• UV Docs: https://docs.astral.sh/uv/