mcp

# 🛰️ MCP with LangChain This project help you to understand MCP and explains how to **trace API calls** made from LangChain's MCP (Model Context Protocol) modules using a simple proxy setup. This will help you to understand the MCP and how the model interacts with it. It includes: - A **proxy server** that intercepts and logs OpenAI API traffic - An **MCP client** that uses LangChain agents to interact with tools .. replace the openai key with your api key .. - A **basic weather tool server** exposed via the MCP protocol --- ## 📦 Requirements Install dependencies using [`uv`](https://github.com/astral-sh/uv) (recommended) or `pip`. --- ## ⚙️ Setup Instructions ### 1. Install `uv` (Recommended) ```bash curl -Ls https://astral.sh/uv/install.sh | sh Verify installation: uv --version ``` 2. Create and Activate a Virtual Environment Using uv: ```bash Copy Edit uv venv source .venv/bin/activate # On Windows: .venv\Scripts\activate ``` Or using python: ```bash Copy Edit python -m venv .venv source .venv/bin/activate ``` Install Dependencies ```bash uv pip install -r requirements.txt # OR pip install -r requirements.txt ``` ### 🏃 How to Run Step 1: Start the Proxy Server In Terminal 1, run: ```bash Copy Edit python proxy.py This will: Start a Flask server on http://localhost:5000 Log every OpenAI API call and response payload ``` Step 2: Run the MCP Client In Terminal 2, run: ```bash Copy Edit python mcpclient.py This will: Start the weather MCP server (mcpweatherserver.py) ``` ✅ What You’ll See In the proxy terminal: full trace of requests/responses to OpenAI In the client terminal: the final agent reply, something like: Response from weather forecast service.. Nice weather today 📁 Project Structure File Description proxy.py HTTP proxy that logs OpenAI API calls mcpclient.py LangChain MCP client that interacts with the weather tool mcpweatherserver.py Minimal MCP tool server using FastMCP and stdio transport requirements.txt Required Python packages 🔍 Why This Project is Useful Gain visibility into OpenAI API usage behind the scenes Understand how MCP-based tools work with LangChain agents Debug and trace tool-based interactions in real time Perfect for learning LangChain + MCP + function calling 🧠 Tech Stack LangChain LangGraph MCP (Model Context Protocol) Flask for proxy OpenAI API (proxied for traceability) 🧪 Example Trace (Proxy Output) ```bash ```` 🚀 Future Ideas Add more MCP tools (e.g., calendar, reminder, calculator) Visualize call chains and LLM decisions Integrate LangSmith tracing for richer insights Happy Tracing! 🧙‍♂️✨ vbnet Copy Edit Let me know if you'd like this saved as a file or want to include code badges, GitHub Actions, or demo screenshots!