MCP Agent

A simple AI agent that uses Model Context Protocol (MCP) to connect LLMs with external tools. This example uses Firecrawl for web scraping capabilities.

What is MCP?

MCP is a protocol that lets AI models interact with external tools and services. Instead of hardcoding tool integrations, MCP provides a standard way for agents to discover and use tools dynamically.

Setup

1. Clone the repo

   git clone https://github.com/yourusername/mcp-agent.git
   cd mcp-agent

2. Install dependencies

   uv sync

3. Set up environment variables

   cp .env.example .env

   Add your API keys to .env:

   OPENAI_API_KEY=your_openai_key
   FIRECRAWL_API_KEY=your_firecrawl_key
   

4. Run the agent

   uv run main.py

Using Different LLM Providers

You can use any OpenAI-compatible endpoint by modifying main.py:

# OpenAI (default)
model = ChatOpenAI(model="gpt-4.1-mini", temperature=0)

# Ollama (local)
model = ChatOpenAI(
    model="llama3",
    base_url="http://localhost:11434/v1",
    api_key="not-needed"
)

# Groq
model = ChatOpenAI(
    model="llama-3.1-70b",
    base_url="https://api.groq.com/openai/v1",
    api_key=os.getenv("GROQ_API_KEY")
)

# NVIDIA NIM
model = ChatOpenAI(
    model="nvidia/nemotron-3-nano-30b-a3b",
    base_url="https://integrate.api.nvidia.com/v1",
    api_key=os.getenv("NVIDIA_API_KEY")
)

Project Structure

mcp-agent/
├── main.py          # Main agent code
├── pyproject.toml   # Dependencies
├── .env             # API keys (not committed)
└── README.md

License

MIT