Privacy Policy

What we collect

Only when you opt in to telemetry (reprompt telemetry on):

• Anonymous install ID — a SHA-256 hash, not reversible to any identity
• 26-dimension PromptDNA feature vectors — numeric counts like word_count, has_code_blocks. These are mathematically impossible to reconstruct into original prompt text
• Score breakdowns — 5 numeric scores on a 0–100 scale
• Bucketed session metadata — duration bucket (e.g., “5–15min”), not exact timestamps
• Client info — client type (cli/extension), locale, reprompt version

Each telemetry payload is under 300 bytes.

What we never collect

Your prompt text. Ever.

All analysis runs locally on your machine. The feature vectors are lossy extractions (regex match counts) that cannot be reversed into text. There is no way to reconstruct what you typed from the numbers we receive.

Share feature

When you use reprompt wrapped --share:

• Uploads: aggregate scores, persona classification, prompt count
• Does NOT upload: individual prompts, prompt text, or raw conversation data
• Share pages expire after 90 days
• Authenticated via HMAC-SHA256 (no passwords or accounts needed)

Data storage

Telemetry data is stored on Cloudflare infrastructure (D1 database, KV store). Share pages are stored in Cloudflare KV with a 90-day TTL. No cross-border transfers of telemetry data beyond Cloudflare’s network.

Your rights

• Opt out anytime: reprompt telemetry off
• No account to delete — we don’t have accounts
• Install ID is anonymous — we can’t identify you even if we wanted to
• Open source: you can audit exactly what’s collected in the source code

Third parties

We do not sell, share, or provide data to any third party. Period.

Contact

Questions or concerns? Open an issue: github.com/reprompt-dev/reprompt/issues