Ollama MCP Server

Local LLM access via the Ollama SDK for MCP-compatible AI applications.

Data last scanned yesterday · Reviewed last month

10 1 dangerous tool13 tools★ 168TypeScriptAGPL-3.0

Overview

Ollama MCP Server exposes the full Ollama SDK as a set of 14 MCP tools, allowing AI assistants like Claude Desktop to interact with local and cloud-hosted models. It features hot-swap architecture for automatic tool discovery and utilizes Zod validation to ensure type safety during interactions. The server supports hybrid model usage, combining local model execution with Ollama Cloud features. It is designed for drop-in integration with clients like Cline, providing capabilities such as web search and content extraction through Ollama Cloud.

Our verdict

This server carries a high-risk score of 34/100, primarily because it is a community-maintained project rather than an official release from the vendor. Additionally, it has not been updated in over 260 days and lacks an editorially reviewed authentication or read-only framework. Due to these maintenance gaps and the lack of professional oversight, users should exercise caution before deploying this in environments where data security or frequent updates are critical.

Provides access to 14 comprehensive Ollama SDK tools
Maintains 96% test coverage for reliable execution
Built with TypeScript for robust type safety
Features a minimal footprint with 5 direct dependencies

Community-maintained project with no official vendor backing
Project has not received a commit in 260 days
Lacks reviewed authentication and read-only access controls
Missing a project homepage link for official reference

Setup

Claude Desktop

{
  "mcpServers": {
    "ollama-mcp": {
      "args": [
        "-y",
        "ollama-mcp"
      ],
      "command": "npx"
    }
  }
}

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.

No read-only mode — running this grants full read/write access.
No built-in authentication — anyone with access to this config can invoke every tool it exposes.

Tools

Tool	Description	Risk
ollama_list	List all available local models	low
ollama_show	Get detailed information about a specific model	low
ollama_pull	Download models from Ollama library	medium
ollama_push	Push models to Ollama library	medium
ollama_copy	Create a copy of an existing model	medium
ollama_delete	Remove models from local storage	high
ollama_create	Create custom models from Modelfile	medium
ollama_ps	List currently running models	low
ollama_generate	Generate text completions	low
ollama_chat	Interactive chat with models (supports tools/functions)	low

Compatibility

Client	Local	Docker	Remote	Read-only
ChatGPT
Claude Desktop
Cursor
VS Code
Windsurf

Frequently asked questions

›What prerequisites are required to use the web search and fetch tools?

You must provide an OLLAMA_API_KEY environment variable to use the ollama_web_search and ollama_web_fetch tools.

›Can I use both local Ollama models and cloud-hosted models simultaneously?

Yes, you can enable hybrid mode by pointing the OLLAMA_HOST environment variable to your local instance while also providing an OLLAMA_API_KEY for cloud access.

›How do I configure a custom Ollama host endpoint?

You can set the OLLAMA_HOST environment variable within your MCP server configuration file to specify a custom endpoint, such as a local or cloud address.

›What is the recommended method for installing this server for Claude Desktop?

You should add the server to your Claude Desktop configuration file using the npx command, which will execute the ollama-mcp package.