Ollama MCP Server

Connect local Ollama models to AI assistants for offline, private AI inference and text generation.

Capabilities

• Model Management: List local models, inspect model metadata, pull new models, and delete unused weights.
• Text Completion & Chat: Execute prompts against locally hosted models like Llama 3, Mistral, and Qwen.
• Embeddings: Generate text embeddings locally without sending data to external APIs.
• Hybrid & Cloud Modes: Connect to Ollama Cloud for web search and retrieval capabilities alongside local execution.

Configuration

Configure your client to launch the server:

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