> Index: https://quebragalho.dev/en/docs/llms.txt

# ZCode

Set it up in the app itself. There is no prompt: an agent cannot reach the app's screens.

- Protocol: chat · responses · messages

## Set it up in the app

- **Where to configure:** Settings › Model Settings › Add Provider, with base `https://api.quebragalho.dev/v1` in the Chat completions format. Saved to `~/.zcode/v2/config.json`.
- **API key:** Pasted into the interface; stored as plain text in the file. ZCode ignores environment variables.

## Caveats

- On each model, set "Max output tokens" to your plan's cap (the default is the model's maximum) and the context to 1M when the model supports it (the default is 200K).
- If the `max` effort returns 400, use `high`.
- Desktop only, by Z.ai. Telemetry includes the host and the model id.

Read it in the browser: https://quebragalho.dev/en/docs/integracoes/zcode
