Local MCP Server with Ollama

A local Python project that connects an Ollama language model to an MCP server. The model can discover the server's tools, decide when to call one, receive the result, and produce a concise final response.

flowchart TD
    U["User question"] --> C["Python MCP client"]
    C --> O["Ollama llama3.2:3b"]
    O --> C
    C --> S["Local MCP server"]
    S --> D["Student JSON data"]

Available tools

Related MCP server: custom-mcp-server

Requirements

• Python 3.11+

• Ollama installed and running

• Ollama model llama3.2:3b

Windows setup

python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
ollama pull llama3.2:3b
python client.py

The client launches server.py automatically through the active Python interpreter. You do not need to start the server in a separate terminal.

Example questions

• What is 10 + 20?

• Tell me about Pranay.

• Where is Ananya studying?

• What is Saranya studying?

Project structure

mcp-server/
├── client.py
├── server.py
├── files/
│   └── students.json
├── tests/
│   └── test_server.py
└── requirements.txt

Tool-result fix

Ollama expects a tool-result message to include the executed tool_name. The client includes this field and passes the available tool definitions into the follow-up call, allowing the model to produce the final response after an MCP tool runs.

Data note

files/students.json is demonstration data. Replace it with fictional or consented data before using this project publicly or in production.