Privacy

Last updated 2 October 2026

VoiceScholr lets you talk live with an AI agent that can hear you and, if the agent uses vision, see your camera or screen. This page says what we collect, where it goes and how long we keep it.

During a call

Your voice, and your camera or screen when the agent uses vision, are streamed live to the AI model so it can answer. We do not record the call audio or video.
The transcript of the call is kept in the server’s memory, so you can see it under the call, and is deleted 10 minutes after the call ends.
If the business that runs the agent has set up a webhook, the transcript and the call’s details (which agent, when the call started and ended, why it ended, and the web page it was on) are also sent to that business’s own systems when the call ends. What they keep is covered by their own privacy policy.
If the business that runs the agent has switched on lead capture, the agent may offer to take your name, email, phone number or what you need. Nothing is saved until you check the details on your screen and tap Confirm. Then the details go to that business, which keeps them until it deletes them. They are stored for it in our database, and they are in the webhook too when the business has set one up. If the business has switched on lead emails, the details are also emailed to it, through Resend.
The agent asks before taking a photo (a snapshot) of your camera or screen. A snapshot shows under the call and is saved privately for the agent’s owner, who can see it in their dashboard. It is deleted after 7 days.
When the agent takes a snapshot, the picture is sent to a second AI model that describes it. The written description, the questions the agent looked up, and the name and email you give when booking a meeting are written to our server logs, which our host deletes after at most 30 days.

Who processes it

Stream: carries the live audio and video between you and the agent.
Google Gemini: the default AI model. The default model runs on Google’s free Gemini API tier, and under Google’s terms for that tier Google may use what it receives (your voice, video and words) to improve its products, and human reviewers may read it. Don’t show or say anything sensitive, such as ID documents, card numbers or medical records, to an agent on the default model. If the agent’s owner uses their own key, it can instead be Alibaba (Qwen), OpenAI, xAI or their own Gemini account. Each provider’s own terms apply to what it receives.
OpenRouter: describes snapshots, and writes agent instructions for owners.
Anam: turns the agent’s voice into a talking face, only when the face is switched on.
Cal.com: receives your name and email when you ask the agent to book a meeting.
Keenable: searches the web when the agent looks something up, and only if the agent’s owner switched web search on. It receives the search words the agent chose, which can include things you said, but not your audio or video. If the owner lets the agent read a whole result page, Keenable also gets that page’s address.
TypeSafe (US): after a call ends, receives the call’s transcript, with email addresses and phone numbers removed first, and your name too when the call knows it, to answer quality questions about it, such as whether you got an answer and how you sounded. TypeSafe’s terms say it does not train its models on what it receives and keeps it only as long as it needs to provide the service.
Supabase: our database and sign-in. Railway: runs our servers. Resend: sends sign-in and feedback emails, and the lead email to a business that has switched it on.

What we store

For every call: which agent, when it started, how long it lasted, and a scrambled visitor id made from your IP address and a browser cookie. This lets us limit free calls. We cannot turn it back into your IP address.
If you sign in: your email address, the agents you build, and any API keys you add. Keys are stored encrypted and never shown again.
If you rate a call: your rating and comment are emailed to us.

Cookies

One cookie (vs_vid) gives your browser a random id for free-call limits. It lasts a year. If you sign in, we also set sign-in cookies. We use no advertising or tracking cookies.

Please don’t share

Don’t show or say card numbers, passwords, one-time codes, or government ID numbers during a call. The agent does not need them.

Your choices

You can turn your microphone and camera off at any time during a call, or hang up.
Signed-in owners can delete their agents and keys from the dashboard.
To delete anything else we hold about you, email us at the address below.

Contact

Questions, or want your data deleted? Email hello@voicescholr.org.