Privacy
What humanjson.org keeps, what it doesn't, and who else handles your data along the way.
Last updated 21 October 2026
- There are no accounts, no cookies and no analytics trackers.
- Your chat, resume, imports and draft stay in your browser. We don't store them.
- To answer you, your messages and attachments go to Anthropic's API.
- What you publish is public. That's the point of it. You can unpublish it at any time.
Who runs this
humanjson.org is run by ZenML GmbH, Schellingstr. 36, 80799 Munich, Germany, which is responsible for the data described here (the "controller" under the GDPR). See the imprint. The human.json spec is open and lives at github.com/htahir1/humanjson. Questions about this notice go there too (see Contact).
What we store
| What | Why | How long |
|---|---|---|
| Your published profile | To serve it at humanjson.org/u/<handle> and its human.json. It contains exactly what you chose to publish. | Until you unpublish it. |
| A hash of your edit token | To check that changes come from you. We never store the token itself, so we can't recover a lost edit link. | Until you unpublish. |
| Which of your accounts link back | When you publish, we fetch each account you list under "other accounts" to check that it links back to your profile, and store which ones do. | Until you republish or unpublish. |
| Feedback notes | Agents that read your file can post a short note about their decision. Only someone with your edit link can read them. | Until you unpublish (we keep the latest 500 per profile). |
| Rate-limit counters | To stop abuse, we count requests per network per day. The counter is keyed by a hash of your network address that changes every day. The address itself is never stored. | Deleted the next night. |
| Usage totals | Plain counts, like page views, which site a visitor came from (for example x.com), profiles created, chat messages and estimated drafting cost. They aren't linked to anyone and use no cookies. | Kept. |
What we don't store
- Your chat with the drafting agent. Messages go through our server to Anthropic and back, and are not written to our database. The conversation is kept in your browser's local storage so you can come back to it. "Start over" deletes it.
- Your resume, LinkedIn export and imports. They're sent along with your next message and then dropped. A LinkedIn export is unpacked in your browser, and only the parts it contains (profile, positions, skills, projects, recommendations) are sent.
- Your draft. It's kept in your browser until you publish it.
- Your edit link. It's kept in your browser so you can get back in, and in your password manager or email if you save it there. The part after
#is never sent to our server in the address. We only see the token when you open your profile, and only as a hash.
Who else handles your data
Cloudflare
The site, the database and the server run on Cloudflare. Cloudflare sees every request, including your network address, as part of serving it. It keeps short-lived request and error logs for us, which we use to fix problems. See Cloudflare's privacy policy.
Anthropic
The drafting agent is Claude, via Anthropic's API. When you send a message, Anthropic receives:
- the message and the earlier ones in the conversation
- your current draft
- any resume or imported data you attached
- the text of pages you linked
It processes them to write the reply. Anthropic's commercial terms say it doesn't train its models on API data. See Anthropic's privacy policy and commercial terms.
GitHub
"+ GitHub" asks GitHub's public API for your profile and repositories directly from your browser, and reads human.json or resume.json from your profile repository if they exist.
Pages you link
When you paste links into the chat, our server fetches those pages, at most five per message, so the agent can read them. Contact details and accounts are never taken from a fetched page.
Fonts and scripts
Pages load fonts from Google Fonts and a few open-source scripts from cdnjs (Cloudflare) and jsDelivr. Those services see your network address when your browser fetches them.
Webmail and password managers
"Email via Gmail" and "Outlook" open a draft in your own webmail. Nothing is sent until you send it. "Save in password manager" uses your browser's own password manager.
What's public
A published profile is public on purpose. Anyone can read it, including people, AI agents and search engines, and they may copy or cache it. Unpublishing removes it from humanjson.org, but we can't recall copies others already made. Leave out anything you wouldn't put on a public web page.
Your choices
- Change it: open your profile with your edit link, edit, and publish the changes.
- Delete it: open your profile, then use ••• → Unpublish. This deletes the profile and its feedback notes straight away.
- Clear your browser: "Start over" deletes the chat and draft kept in your browser. Clearing the site's data in your browser removes everything, including the saved edit link.
- Lost your edit link? We can't recover it, because we only keep a hash. Get in touch and we'll help you prove the profile is yours and remove or reset it.
Contact
Open an issue at github.com/htahir1/humanjson/issues. For something private, like a lost edit link or a profile published about you without your consent, open an issue saying only that you'd like to be contacted. Don't put personal details in it, and we'll reply privately.
Changes
If this notice changes, the date at the top changes too. Big changes will be announced on the GitHub repo.