Ollama MCP Server

A lightweight Model Context Protocol (MCP) server that exposes Ollama Cloud models as tools to MCP clients such as Claude Code.

This server not only lets you chat with Ollama Cloud models — it also makes it easy to switch between the best model for each task: coding, math, reasoning, agentic workflows, and more.

Features

• List models – Discover all models available on your Ollama Cloud instance.

• Get / set default model – Switch the active model at runtime to match the task.

• Chat – Send prompts to any Ollama Cloud model, with optional system messages and per-request model overrides.

• Web search – Search the web through Ollama's hosted web search API.

• Web fetch – Fetch and extract a webpage's content through Ollama's hosted web fetch API.

• Task-aware model guide – Built-in recommendations for choosing the right model for coding, math, reasoning, and other workloads.

Related MCP server: mcp-ollama-python

Requirements

• Python 3.10+

• An Ollama Cloud account with an API key.

Installation

1. Clone the repository:

   git clone https://github.com/chakkritt/ollama-mcp.git
   cd ollama-mcp

2. Create a virtual environment (recommended):

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

Configuration

Copy .env.example to .env and fill in your API key:

cp .env.example .env

Then edit .env and replace the placeholder values with your own:

OLLAMA_HOST=https://ollama.com
OLLAMA_API_KEY=YOUR_OLLAMA_API_KEY
DEFAULT_MODEL=gpt-oss:120b

• OLLAMA_HOST – Base URL of your Ollama Cloud instance.

• OLLAMA_API_KEY – Your Ollama Cloud API key.

• DEFAULT_MODEL – Fallback model used when a chat request does not specify one.

You can override the model on every chat request, so the DEFAULT_MODEL is just a sensible starting point.

Model selection guide

Use this guide to pick the best Ollama Cloud model for your task. Switch models anytime with the set_current_model tool.

Coding & software engineering

These models excel at writing, debugging, and reasoning about code.

Recommended default for coding: glm-5.2 or kimi-k2.7-code.

Math, reasoning & problem solving

Recommended default for math: deepseek-v4-pro.

General-purpose & agentic tasks

Multimodal & vision

Fast / cost-efficient tasks

Running the server

python server.py

By default, the server runs as an MCP server over stdio, which is the standard transport for most MCP clients.

Available tools

Example usage with Claude Code

Add the server to your Claude Code MCP configuration:

{
  "mcpServers": {
    "ollama": {
      "command": "python",
      "args": ["/absolute/path/to/server.py"],
      "env": {
        "OLLAMA_HOST": "https://ollama.com",
        "OLLAMA_API_KEY": "YOUR_OLLAMA_API_KEY",
        "DEFAULT_MODEL": "gpt-oss:120b"
      }
    }
  }
}

Then you can ask Claude to:

• "List available Ollama models"

• "Switch to the coding model" — the guide recommends glm-5.2 or kimi-k2.7-code.

• "Use the math model to solve this problem" — the guide recommends deepseek-v4-pro.

• "Chat with glm-5.2 and refactor this function"

• "Set the default model to deepseek-v4-pro"

• "Think through this problem with deepseek-v4-pro" — calls chat with think: true and returns the reasoning trace in thinking.

• "Search the web for the latest news on …" — uses web_search.

• "Fetch the page at https://… and summarize it" — uses web_fetch.

Quick task-to-model cheatsheet

Project structure

ollama-mcp/
├── .env.example      # Environment variable template
├── .gitignore        # Files ignored by Git
├── LICENSE           # MIT license
├── pyproject.toml    # Project metadata
├── README.md         # This file
├── requirements.txt  # Python dependencies
└── server.py         # MCP server implementation

Dependencies

• fastmcp – Framework for building MCP servers in Python.

• ollama – Official Ollama Python client.

• httpx – HTTP client for the Ollama hosted web search/fetch APIs.

• python-dotenv – Load environment variables from .env.

Keeping model recommendations up to date

Ollama Cloud's model catalog changes frequently. To refresh this guide:

1. Visit https://ollama.com/search?c=cloud.

2. Update the Model selection guide and Quick task-to-model cheatsheet sections with new releases or benchmarks.

3. Adjust your DEFAULT_MODEL and set_current_model calls to match your current workload.

License

Apache License 2.0