Ollama MCP Wrapper

A Python-based Model Context Protocol (MCP) server wrapping local Ollama capabilities. This allows AI agents to interface with local language models, query tags, generate text, and execute tool-calling sequences under strict prompt injection static sanitization.

File Structure

├── .agents/             # Agent rules & skills
├── src/
│   └── ollama_wrapper/  # Package source
│       ├── __init__.py
│       ├── config.py    # Configuration defaults
│       ├── security.py  # Prompt injection scanner
│       └── server.py    # FastMCP server
├── tests/               # Pytest testing suites
└── pyproject.toml       # Python dependencies

Related MCP server: Ollama MCP Server

Setup & Running

Requirements

• uv for Python environment management.

• Ollama running locally (defaults to http://localhost:11434).

Running the Server

To run the server locally via standard I/O (STDIO):

$env:PYTHONPATH="src"
uv run python -m ollama_wrapper.server

Running Tests

To run the unit tests:

$env:PYTHONPATH="src"
uv run pytest