Pick a starting point, write a message, or drop files.
Configure system prompt, global sampling overrides, display width, and paste behavior.
Configure a provider once, then add and customize multiple models under it.
Persist conversations and assets across devices using your private GitHub repository.
Create reusable prompt snippets invoked by @name in a message.
Connect local or remote MCP servers over Streamable HTTP or SSE.
The assistant is requesting to run an external tool.
Global proxy to bypass browser CORS for LLM providers and remote MCP servers.