Connect to a model provider

Bring the coding agents you already use, or run a hosted model or your own private local model, all from Settings → Providers.

Cue for Mac Providers settings with Claude Code, Codex, Antigravity, OpenRouter, Local models, and Apple Intelligence in two groups.

Choose a provider

The difference is who runs the agent: a CLI you installed, or Cue itself.

Installed agents

Cue starts the CLI in your login shell, so it has the same PATH, keys and sign-in as your terminal. Your past sessions with it appear in History and resume in Cue.

Models through Cue

Cue runs the agent itself. The model writes short scripts that call Cue’s tools for files, search, the web, screenshots and documents.

Set up a provider

Each takes a one-time setup. After that, pick it from the agent menu in any chat.

Claude Code, Codex, or Antigravity

  1. Install the CLI and sign in from your terminal: claude, codex or agy.
  2. Pick it from the agent menu. Cue finds it on your shell’s PATH.
  3. Optionally set its model, thinking and access in Settings → Providers.

Apple Intelligence

  1. Turn on Apple Intelligence in macOS System Settings on an eligible Mac.
  2. Let its model finish downloading. Until then, the agent menu greys it out and says why.
  3. Pick Apple Intelligence from the agent menu.

OpenRouter

  1. In Settings → Providers → OpenRouter, choose Connect OpenRouter. You sign in on openrouter.ai and Cue keeps the key it issues. Have a key already? Choose Use an API key instead… or set OPENROUTER_API_KEY in your shell.
  2. Pick a model, or choose All Models… to search every model, free or paid. Paid requests use the connected account’s credits.
Cue’s OpenRouter settings with account connection and model browsing controls.

Local models

  1. Open Settings → Providers → Local models.
  2. Cue on this Mac: download a recommended GGUF model or import your own. Cue runs it with its bundled llama.cpp, so your chats never leave the Mac.
  3. Existing llama.cpp server: enter its address and any API key, then Save & Check to list its models.
Cue’s Local models settings with Cue on this Mac and Existing llama.cpp server.