genjutsu collects nothing.

The plugin

genjutsu is a set of Markdown instructions and local Python scripts. Claude reads them and acts on your machine, inside the session you started. Nothing is transmitted anywhere.

• No telemetry, no usage reporting, no crash reporting, no phoning home. No script shipped with the plugin makes a network request.
• No account, no sign-up, no licence key, no identifier of any kind.
• It reads and writes files in the project you point it at, and only there. Nothing is copied out of it.

One honest caveat. The two skills declare WebSearch among their allowed tools, so Claude may run a web search when a task genuinely calls for it. That request is made by Claude, under Anthropic's policy, not by genjutsu. What ends up in the query is Claude's call, and not something genjutsu controls.

This site

• No analytics, no tracking pixel, no fingerprinting, no cookies.
• One entry in localStorage, genjutsu:theme, remembering whether you chose light or dark. It never leaves your browser and clearing site data removes it. Never setting it is fine too: the site follows your operating system instead.
• No third-party request from your browser. The star count and the version number are fetched from GitHub when the site is built, so they are already in the page by the time you load it.
• Static files on a hosting provider, which keeps the ordinary web server access log any host keeps: IP address, user agent, requested URL, timestamp. It is not read by genjutsu, not joined to anything, and not used to build a profile.

Where other policies take over

genjutsu runs inside other people's software, and those parties have their own terms.

• Installing the plugin fetches it from GitHub, so GitHub sees that request. Their privacy policy applies.
• Running it means you are in a conversation with Claude, so Anthropic's privacy policy applies to that conversation.

Neither of those is something genjutsu mediates or can see.

Changes, and asking

This page is versioned with the site, so every edit it has ever had is in the repository, with its date and its diff. There is no separate archive to trust.

Anything unclear, or a claim here you want substantiated, goes inthe issue tracker. The source is public: every statement above can be checked against it.