Privacy Policy

Status

Plain-language summary

What we collect

• Account data: email address (for license delivery and billing receipts) and a Polar customer ID. We never see or store your card details.
• Voice transcripts: when you speak to Darvy, your audio is sent to a speech-to-text provider and your text is sent to a large-language-model provider. Audio is not retained by us after transcription. Transcripts may be stored locally on your machine if you enable history; they are not sent to our servers by default.
• Diagnostics: anonymous error reports if Darvy crashes (no user content, no audio).

What we don't do

• We don't sell your data to anyone, ever.
• We don't train AI models on your voice or your transcripts.
• We don't keep audio recordings server-side.
• We don't read your emails, documents, or files. Darvy reads them on your machine, in the moment, to do the task you asked for.

Categories of third parties we use

• Payment processor — PCI-compliant merchant of record (currently Polar).
• Speech-to-text provider — transcribes your microphone audio.
• Large-language-model provider — interprets transcribed text to decide what action to take.
• Text-to-speech provider — generates Darvy's spoken replies.
• Web hosting — serves this website.
• Crash reporting — anonymous diagnostics when Darvy crashes.

The current vendor in each category is available on request — email support@darvy.ai with subject "Sub-processor list" and we'll send the up-to-date list. We list categories rather than vendor names because our infrastructure choices change over time; the categories of data handled do not.

How to ask for your data, or delete it

Email support@darvy.ai with the subject "Data request". We respond within 30 days (usually within 48 hours).

Where the binding version will live

Once finalised, the canonical privacy policy will be committed to the project repo at legal/privacy.md and rendered here verbatim. Until then, treat this page as our public commitment.