Ollama MCP

What it does

Ollama MCP connects your AI assistant to a local Ollama instance, providing full programmatic control over local LLMs. It allows the AI to manage models, generate completions, and create embeddings without leaving the chat interface.

Tools

• ollama_list: List available local models.
• ollama_show: Get detailed model info.
• ollama_pull: Download models from the Ollama library.
• ollama_push: Push models to the library.
• ollama_copy: Duplicate existing models.
• ollama_delete: Remove models from storage.
• ollama_create: Create custom models via Modelfile.
• ollama_ps: List currently running models.
• ollama_generate: Generate text completions.
• ollama_chat: Interactive chat with local models.
• ollama_embed: Generate text embeddings.
• ollama_web_search: Search the web via Ollama Cloud (requires API key).
• ollama_web_fetch: Fetch web page content via Ollama Cloud (requires API key).

Installation

Add to claude_desktop_config.json:

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

Supported hosts

• Claude Desktop
• Windsurf
• Cursor
• Cline