mcp-client-server-example

# MCP + Ollama Local Tool Calling Example This project demonstrates how a local AI agent can **understand user queries** and **automatically call Python functions** using: - Model Context Protocol (**MCP**) - **Ollama** for running a local LLM (e.g., Llama3) - **Python** MCP Client and Server --- ## 🔗 Sequence Diagram ```mermaid sequenceDiagram participant User participant MCP_Client participant Ollama_LLM participant MCP_Server User->>MCP_Client: 1) User types: "What is 5 + 8?" MCP_Client->>Ollama_LLM: 2) Send available tools + user query Ollama_LLM->>Ollama_LLM: 3) Understand query & tool descriptions Ollama_LLM->>Ollama_LLM: 4) Select tool: add(a=5, b=8) Ollama_LLM->>MCP_Client: 5) Return tool_call MCP_Client->>MCP_Server: 6) Execute add(a=5, b=8) MCP_Server-->>MCP_Client: 7) Return result: 13 MCP_Client-->>User: 8) Show final answer: 13 ``` --- ## 📚 Project Structure ``` . ├── math_server.py # MCP Server exposing add() and multiply() tools ├── ollama_client.py # MCP Client interacting with Ollama ├── README.md # Project documentation ``` --- ## 🛠️ Setup Instructions ### 1. Install Requirements ```bash pip install "mcp[cli] @ git+https://github.com/awslabs/mcp.git" openai==0.28 httpx ``` Make sure you have **Ollama installed** and running. ### 2. Pull or run an LLM model ```bash ollama run llama3 ``` (Ensure the model you run supports tool calling.) ### 3. Run the MCP Server ```bash python math_server.py ``` The server exposes two simple tools: - `add(a: int, b: int) -> int` - `multiply(a: int, b: int) -> int` ### 4. Run the MCP Client ```bash python ollama_client.py math_server.py ``` ### 5. Interact! Example queries: ``` Query: What is 5 + 8? Response: 13 Query: Multiply 7 and 9 Response: 63 ``` The MCP client sends the query and available tools to Ollama. The LLM internally decides which tool to use based on the tool descriptions and user intent. --- ## 🚀 How It Works - MCP Client lists available tools. - Sends tools + user query to Ollama LLM. - LLM reasons about the best matching tool. - LLM generates a `tool_call`. - MCP Client invokes the function via the MCP Server. - Final result is returned and displayed. ✅ No manual hardcoding! ✅ Everything runs locally! ✅ Fully autonomous! --- ## 📢 Why This Matters This pattern enables building **smart local AI agents** that: - Understand user intent - Dynamically select the correct actions - Operate fully offline and locally It opens doors for: - Autonomous developers - Local intelligent assistants - Secure AI workflows --- ## 🏷️ Hashtags for Sharing ```text #MCP #ModelContextProtocol #Ollama #LocalLLM #FunctionCalling #Python #AI #DeveloperTools #AIEngineering #AutonomousAgents ``` --- ## 🙌 Credits - [Model Context Protocol](https://modelcontextprotocol.io) - [Ollama](https://ollama.com) --- > "Smarter AI agents start with understanding how they think!" --- > **Next Steps:** Add Streamlit UI or Dockerize this project 🚀