- When
- When an AI feature runs: reading a résumé you add, fitting it to a job, answering a form’s questions (and translating them when the app is in Russian), rating found jobs (10 per role a day at most, by default), suggesting more job titles once for a new role, an interview prep, a follow-up email, a look at an interview transcript you add and, only if AI email sorting is on, your job mail. Also when you connect it and pick a model.
- Where it goes
- Only the one you connect. An OpenAI key:
api.openai.com. An Anthropic key: api.anthropic.com. Your ChatGPT plan: you sign in on auth.openai.com in your browser, then requests go to api.openai.com. Your Claude plan: the app runs Claude Code, Anthropic’s own program on your computer, and Claude Code sends the request to Anthropic. Ollama: localhost:11434, on your computer, unless you type another address.
- What is sent
- What the feature needs: the job’s text, your résumé, a form’s questions with the About you answers they need (name, city, work authorization, salary expectation, links), a role’s job titles, a job’s notes and interview transcripts for its follow-up, and email only with AI email sorting on (its subject, sender, date and text). Your key or sign-in goes with each request, to that provider only. Signing in with ChatGPT also gives OpenAI a random id this copy of the app made, which OpenAI’s sign-in asks for. Install Claude Code, if you press it, downloads Anthropic’s installer from
claude.ai.
- Never sent
- The street, apartment and postal code you save in About you (the AI sees an address only if your résumé has one), your mail password, anything to Weavenly. Your key goes nowhere but its own provider, and with a ChatGPT or Claude plan the app never sees your password. That provider’s own policy applies to what it gets.
- Can you turn it off
- Yes. With no AI connected, nothing goes to any provider. AI email sorting stays off until you turn it on, and AI ratings per role per day set to 0 turns the ratings off.