Skip to main content
Kimi uses local kimi. Open Settings → CLI configuration → Kimi CLI to detect the version, install, and configure credentials.

​
How to connect

  1. Use local config: read ~/.kimi-code/config.toml directly. Override the config directory with $KIMI_CODE_HOME
  2. Add provider: name, base URL, API key, model. Enabling it writes that config.toml (backed up first as config.toml.bak)
  3. You can override the kimi executable path
  4. If it is not installed, click Install on that page, or install with the terminal commands below
Config stays on this computer. The app does not upload keys to a CC GUI server.

​
Add a provider

Click Add and fill in:
  • Name, base URL, API key, model
  • Provider type: Kimi Code (official) / OpenAI-compatible / Anthropic-compatible
  • Optional: max context length, display name, notes
Quick presets: Kimi Code (official), Moonshot Open Platform. You can fetch the model list from the remote. The hosted provider catalog is ~/.ccgui/config.json. Click Use local config.toml and that file is left unchanged.

​
Install

macOS / Linux:
Windows PowerShell:
Alternative:
In-app Install previews the fallback npm install -g @moonshot-ai/kimi-code. Docs: Kimi Code

​
Capability differences

Compared with Claude Code:
  • No plan mode
  • Does not use desktop MCP
  • Supports session continue and image input
  • Stop ends the kimi child process
  • History is isolated per runtime

​
When detect fails

See Detect and install.

​
Next steps