Privacy
This page describes the hosted instance at app.thinkclear.xyz. A self-hosted copy stores the same things in a database you control, and none of it reaches anyone here.
What this marketing page collects
Nothing. thinkclear.xyz is a set of static files with no analytics script, no advertising pixel, no third-party embeds, and no cookies of its own. Fonts are served from this origin rather than a font host, so loading the page does not tell anyone else that you did. The hosting provider keeps ordinary request logs, as every web server does.
What the app stores
An account is an email address, a display name, and either a password hash or a link to the identity provider you signed in with. Everything else in the database is content you made: your mindmaps and their topics, the markdown notes on those topics, your conversations with the assistant, your layout preference, and the list of agent clients you have granted access to.
Every one of those rows is scoped to the account that owns it, and every read is filtered by owner before it is filtered by anything else. There is no admin view that browses other people's maps.
What leaves the server
- Messages you send to the assistant, and the parts of a mindmap it reads to answer them, go to the configured model provider through LLM Gateway. That is the one place your content is handled by a third party, and it happens only when you use the assistant.
- If you sign in with Google, Google learns that you signed in here — the ordinary consequence of using it as an identity provider. Signing in with an email and password involves no third party at all.
- Nothing is sold, and nothing is shared for advertising. There is no advertising.
Cookies and tokens
The app sets one session cookie, first-party and host-only on app.thinkclear.xyz. It is not readable by this page, and it is deliberately not widened to cover the whole domain. There are no tracking cookies, so there is no consent banner to dismiss.
An agent connected over MCP holds an OAuth access token instead of a cookie, bound to the scopes you approved on a consent screen. A token granted read access is served a tool list with no way to edit anything — the restriction is that the tools are not registered, not that the calls are refused. Every grant is listed in the account screen and can be revoked there, which takes the agent's access away immediately.
Keeping and deleting
Content is kept until you delete it. Deleting a mindmap or a conversation removes the document; deleting your account from the account screen removes the account and the content that hangs off it. Backups of the database roll over on their own schedule, so a deletion can survive in a backup for a short window after it has left the live database.
If you would rather none of this were somebody else's problem, the source is AGPL-3.0 and the deployment guide is public: run it yourself and the only privacy policy that applies is your own.