Model-Agnostic
Rook isn’t tied to one AI vendor. It speaks the Model Context Protocol (MCP), so any client that does will work — and its agent layer is routed through LiteLLM, giving you a hundred-odd providers, local or cloud.
One plugin, any mind.
Clients that work
Section titled “Clients that work”| Client | Support |
|---|---|
| Claude desktop app (Chat, Cowork, Code) | Recommended · tools + skills |
| ChatGPT desktop app (Codex) | Fully supported · tools + skills |
| Claude Code command line | Tools + skills |
| Cursor | MCP tools |
| Windsurf | MCP tools |
| Ollama | Local · agent backend |
| LM Studio | Local · agent backend |
| …and 100+ more | via LiteLLM |
Why it matters
Section titled “Why it matters”- Use what you already have. No new account or vendor lock-in — point your current assistant at Rook.
- Run local if you want. Ollama and LM Studio keep everything on your machine.
- Switch freely. Change models without changing how you work.
You choose the mind; Rook provides the hands.
Related
Section titled “Related”- Many Providers — the same openness for image, video, and 3D generation
- Install Rook — connecting your client