Frequently asked questions
Everything about the personal AI mediator, the Ideas and Facts modes, privacy and pricing.
What is an AI mediator?
It's a layer in front of the AI that turns your everyday wording into a precise request. You say what you need in your own words, it asks one or two simple questions, and it writes the prompt for you.
How is Facts mode different from a normal chat?
In Facts mode the answer is grounded in sources you provide — uploaded documents, pasted text or web search. Quotes are checked against the source and arithmetic is computed exactly, so any claim that isn't backed is marked unverified.
Are my documents safe?
We store only the extracted text needed to answer, never the original file, and each source is tied to a single conversation. You can export or erase all of your data at any time.
Is it free?
Yes, you can start for free. A monthly message quota is included, and paid plans raise the limits when you need more.
Which languages are supported?
The interface ships in four languages — English, German, Russian and Ukrainian — and the AI can chat in many more.
How is this different from ChatGPT?
Instead of a blank box, you get a guided cycle: rough input, a couple of clarifying questions, a clear before and after, and a source-grounded Facts mode for documents and the web.
Do I need to know how to write prompts?
No. That's the whole point — you write however you can, and the mediator turns it into a well-formed request.
What is the difference between Ideas and Facts mode?
Ideas mode helps you think and create from an open brief; Facts mode answers strictly from sources you supply — your documents, pasted text, or web results — with verifiable citations.
Does the AI make things up?
Any language model can hallucinate. Facts mode is built to prevent it: answers are grounded in your sources and every quoted claim is verified against them, so an unbacked statement is flagged as unverified.
Can I combine my own files with web search?
Yes. In Facts mode you can attach documents, paste raw text, and enable web search at the same time. The assistant prefers your sources and reaches for the web only when they don't answer.
Which AI models can I use?
Anthropic (Claude), OpenAI, and Google (Gemini) are all supported. The model is pinned when a conversation starts, so later changes never alter past chats.
How do citations work?
When the assistant quotes a source, the exact quote must actually appear in that source or it is shown as unverified — so you can always check a claim against the original.