# Token Tunnel: use your Claude, ChatGPT, Gemini, or Grok plan

> Token Tunnel connects the AI subscriptions you already pay for, so chats and workflows run on them with no API keys and no per-token charges.

Jul 19, 2026. Topics: Models, Platform. https://www.generalinput.com/changelog/2026-07-19-token-tunnel

[Token Tunnel](/token-tunnel) is available to every workspace. Install General Input Desktop, sign in to the AI subscriptions you already have (Claude, ChatGPT, Gemini, or Grok), and those models appear in General Input as your own. Chats and agent workflows run on them without an API key, usage isn't billed per token, and your provider logins live only on your computer. Learn more about [running local and personal models](/blog/what-is-a-local-ai-model).

### How it works

- **A menu bar app.** One universal macOS build for Apple Silicon and Intel. Sign in to General Input with a device code, then to each provider with its own login. The app auto-connects and updates itself.
- **The tunnel.** The app keeps an outbound connection to our gateway. When a run uses one of your models, the request is relayed to your computer, sent to the provider on your subscription, and streamed back.
- **Model sync.** Any model in our catalog that your subscription serves appears automatically, including models we don't host, like Claude Fable 5.
- **Explicit offline handling.** If your computer is off, chat says so immediately and never silently switches to a paid model. Agent workflows choose a policy: skip the run after retrying for 15 minutes, or fall back to a hosted model you pick.
- **Clear errors.** Rate limits, expired logins, and provider outages are labeled as such, and each device serves 10 simultaneous requests by default.