Privacy policy
Last updated 17 November 2026
Harold is a beta run by Tariq Khalif. This page explains what Harold collects, why, who helps process it and how to have it deleted.
What Harold collects
- Your Telegram user id, name and username.
- The messages you send Harold, including voice notes and their transcripts.
- Drafts, rendered slides, schedules and the facts Harold learns about your work.
- If you connect them: your Instagram access token, account id and post insights, and your GitHub username, token and recent public activity.
- A session cookie when you sign in to the web app.
Why
Only to run Harold for you: write and render posts, publish what you approve, report results and remember your preferences. Harold has no ads, does not sell data and does not use it to train models.
Who processes it
- Cloudflare hosts Harold and stores its data, and transcribes voice notes.
- Anthropic (Claude) generates Harold's replies and drafts from your messages.
- Telegram delivers messages between you and Harold.
- Meta receives posts you approve for Instagram and returns their insights.
- GitHub provides your activity if you connect it.
Security
Instagram and GitHub tokens are encrypted at rest and never shown back. Web sessions can be ended on every device with /signout.
Keeping and deleting data
Data is kept while you use Harold. To delete it, ask in the Harold chat; during the beta this is done by hand and finished within 30 days.
Changes
If this policy changes, the date above changes with it, and Harold will tell you in the chat about anything significant.