Providers & BYOK
Providers & BYOK
Calca talks to AI providers directly from the app — there is no Calca account
and no Calca-side proxy holding your key. Your key is stored in local storage
and every request goes from your browser (or the desktop app) to the provider
endpoint you configured.
Supported provider types
| Type | Use it for | Fields |
|---|---|---|
| OpenAI-Compatible | OpenRouter, OpenAI, Groq, Together, LM Studio, vLLM, any /v1/chat/completions endpoint |
API key, base URL, model id |
| Anthropic | Claude models | API key, model id |
Recommended setups
OpenRouter (easiest, one key for many models)
- Base URL:
https://openrouter.ai/api/v1 - Model:
google/gemini-2.5-flashis a good cost/quality floor; heavier models
follow your budget.
Anthropic
- Model:
claude-sonnet-4-5or your preferred Claude.
Local (free, offline)
- Run LM Studio, load any model, start the local server.
- Base URL:
http://localhost:1234/v1, API key: leave empty. - Great for iterating without spending tokens; expect weaker layouts from
small local models.
Two-provider setup
Settings lets you keep a primary provider and configure additional
providers to switch between. If generation fails or a model is rate-limited,
swap the active provider/model from the toolbar — the cheapest fallback that
still holds multi-component layouts is google/gemini-2.5-flash on OpenRouter.
Rough cost guide: generating a batch of design variations on
gemini-2.5-flash costs well under a cent. Sessions of everyday experimentation
stay under a dollar unless you deliberately pick frontier models.
