This server wraps the complete Ollama SDK, giving Claude Desktop and other MCP clients direct access to your local LLM models. You get 14 tools covering the full lifecycle: pull/push/delete models, generate completions, run chat sessions with function calling support, create embeddings, and manage running processes. It also connects to Ollama Cloud for web search and content fetching when you provide an API key. The retry logic handles rate limits intelligently. If you're already running Ollama locally and want Claude to orchestrate your models or need hybrid local/cloud workflows, this bridges that gap cleanly.
claude mcp add --transport stdio rawveg-ollama-mcp uvx ollama-mcp