Privacy policy
Effective 2026-08-04 · updated 2026-08-22 (usage-insights and update-check descriptions corrected against the code) · changes are dated on this page
The short version
no_human is a coding agent that runs on your own machine. We never receive your code, your prompts, or your tickets. The personal data we hold is what you hand us directly: an email address on the waitlist or an account sign-in. If you switch on the optional usage-insights setting in the app, we additionally receive anonymous usage events and masked recordings of the app's own interface — described precisely below, off until you turn them on, and never containing your code.
What we collect, and why
- Waitlist: your email address, the plan you selected, a timestamp, and your browser's user-agent string — stored on AWS and used only to contact you about availability.
- Account: the email address from your Google sign-in or email code, held in AWS Cognito and used to operate your account.
- Support: any email you send us, kept as ordinary correspondence.
This website sets no analytics trackers and no advertising cookies. Sign-in uses only the cookies and tokens it strictly needs to work.
Usage insights in the app (optional)
no_human includes a single usage insights setting. It lives in the board served by the local server, which the desktop app starts for you and which a command-line install also runs — so this section applies to both. It is off by default; nothing in this section happens unless you turn it on, and turning it off stops it. When on, two things are sent to us:
- Usage events: the app's own events are a short list — app started, task created, task completed (status, duration bucket, attempt count), task failed (error category), approve clicked, feature used, and screen viewed. Of those, only screen viewed is sent from the app window; the other six are sent by the local server, and on a stock install they go nowhere at all — that channel needs a configured endpoint and ships without one, so it is inert until someone sets it. Server-sent events carry a random installation ID and the app version; window events carry the app version and the analytics provider's own anonymous device identifier. The provider's library also captures its built-in interaction and page-view events, which can include the text of a control you clicked. We intend to disable that and have not yet; and until we have, treat this list as the app's events rather than as everything sent.
- Interface replays: a recording of the no_human window only — screens visited and controls used — so we can see where the product confuses people. It cannot capture your desktop, other applications, or anything outside the app's own window. Inside the window, everything you type is masked, and every surface that renders your content — code, diffs, task titles and descriptions, repository names, file paths, logs and pull-request text — is marked no-capture, so replays show that a screen was used rather than what your work said. That marking is applied surface by surface, and a source-level check flags a component that renders a known task field with no masking annotation anywhere in the file — a backstop, not a proof: it does not catch a field destructured out of the object first, and it does not look at where in a file the annotation sits.
The random installation ID is minted by the app and is not derived from your account, hardware, or network; we do not attempt to link it to you. Usage events flow to our own AWS storage. Interface replays and events are processed by PostHog, Inc. (US cloud) as our processor. We use this data to improve the product and for no other purpose; retention is bounded (our event store deletes after 90 days) and you can have it deleted sooner on request.
What we never see
The no_human app runs locally. Your repositories, code, tasks and prompts go, in the app's own traffic, only to the AI provider you configure, the git forge you point it at and any tracker, chat or CI integration you enable, under your own credentials — that traffic never passes through us, with usage insights on or off. The app periodically checks for updates — the desktop build against GitHub Releases, the Python package against PyPI. Those are third-party hosts, not ours, so the request does not reach us; it carries no personal data beyond what any web request includes. The app also runs your project's own install, build and test commands, and its agents work in an ordinary shell on your machine, so while they work they can reach package registries and other hosts they choose — from your machine, never through us.
The Jira app
Our Jira app runs entirely inside Atlassian's platform, acts only with the permissions of the person clicking, and cannot reach any host outside Atlassian. It stores one thing: the label name you configure. It stores no personal data.
Cloud features
Hosted execution and team features are not yet live. Before any such feature processes customer data, this policy will be updated to say exactly what it touches.
Who processes data for us
AWS (waitlist, accounts, usage events, and transactional email), Cloudflare (serving this website), Atlassian (the Jira app), Google (optional sign-in), and PostHog, Inc. (optional usage insights, US cloud). We do not sell or share personal data with anyone else.
Retention and deletion
Waitlist and account data are kept until you ask us to remove them. Email eyalgolan96@gmail.com and we will delete your data and confirm.