READMEmaster

MCP + FastAPI + Ollama 🧠🚀

Name: llm-mcp-fastapi
Author: ahmadalsharef994

A minimal FastAPI backend that exposes tools to local LLMs using the Model Context Protocol (MCP). Connect any MCP-compatible LLM (LLaMA 3 via Ollama, Claude, etc.) to your tools with a clean REST API.

✨ Features

FastAPI backend with MCP-compatible endpoints
Tool listing: GET /mcp/tools/list
Tool invocation: POST /mcp/tools/invoke
Example tool: get_weather(city)
Works with Ollama and any LLM supporting tool calling
Easy to extend — add new tools in minutes

🏗️ How It Works

sequenceDiagram
    participant LLM as 🦙 Ollama LLM
    participant API as FastAPI MCP Server
    participant Tool as Tool Implementation

    LLM->>API: GET /mcp/tools/list
    API-->>LLM: [{name, description, parameters}]

    LLM->>API: POST /mcp/tools/invoke\n{tool: "get_weather", args: {city: "Berlin"}}
    API->>Tool: execute get_weather("Berlin")
    Tool-->>API: {"temperature": "12°C", "condition": "cloudy"}
    API-->>LLM: Tool result
    LLM->>LLM: Generate final response using tool result

🚀 Quick Start

git clone https://github.com/ahmadalsharef994/llm-mcp-fastapi.git
cd llm-mcp-fastapi

python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

# Start Ollama (in a separate terminal)
ollama serve
ollama pull llama3.2

# Start the MCP server
uvicorn app.main:app --reload

🔌 API

List available tools

curl http://localhost:8000/mcp/tools/list

[
  {
    "name": "get_weather",
    "description": "Get current weather for a city",
    "parameters": {
      "city": {"type": "string", "required": true}
    }
  }
]

Invoke a tool

curl -X POST http://localhost:8000/mcp/tools/invoke \
  -H "Content-Type: application/json" \
  -d '{"tool": "get_weather", "args": {"city": "Berlin"}}'

Chat with tool-calling

curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "What is the weather in Berlin?"}'

🛠️ Adding New Tools

# app/tools/my_tool.py
def my_tool(param: str) -> dict:
    return {"result": f"processed {param}"}

# Register in app/main.py
tools.register("my_tool", my_tool, description="Does something useful")

📁 Project Structure

app/
├── main.py           # FastAPI app + MCP endpoints
├── tools/            # Tool implementations
│   └── weather.py    # Example weather tool
└── ollama_config/    # Ollama model settings

📄 License

MIT