Ask AI providers
Luora hosts no AI: Ask answers from built-in text, or a model you bring.
Open Connect AI in the app's top bar.
| Option | What answers | What leaves your machine |
|---|---|---|
| None (default) | Luora's built-in text | Nothing |
| Local model | Ollama, LM Studio or any OpenAI-compatible server you run | The prompt goes from your browser to that endpoint (normally localhost) |
| Own API key | Anthropic, OpenAI or OpenRouter | The prompt and key go from your browser straight to that provider |
| Your Claude | Your own Claude Code with your Claude plan | Local Luora only (Run Luora locally) |
What is sent
Only when you ask, and only for the selected piece or calls: their source lines, the connected pieces, open findings and your question, plus a fixed system prompt that keeps the model defensive (explain, list what to check, write a test of intended behaviour, never exploits).
Local model on luorasec.xyz
The browser calls your local server directly, so it must allow the site. With Ollama:
launchctl setenv OLLAMA_ORIGINS "https://www.luorasec.xyz" # macOS, then restart Ollama ollama pull llama3.1
Then pick Local model, endpoint http://localhost:11434/v1, model llama3.1, and press Test connection. In LM Studio, turn on Enable CORS in the server settings.
Your API key is kept only in this browser's localStorage and sent only to the provider you picked. Use a key you can revoke, with a spend limit.