A note on this build. Wren is a reference / demo build made for cshell's calibration sweep. It is not a commercial service, and it is not affiliated with Anthropic. This policy is written to read like the real thing and to describe exactly how a demo like this actually handles data — no more, no less.
This policy explains what information Wren collects when you use the web app, why we collect it, and the choices you have. We've tried to write it in plain language. Where we simplify, the intent is honesty, not evasion — if something here is unclear, get in touch and we'll explain it directly.
We keep collection narrow — only what's needed to make Wren work.
We do not ask for and do not want sensitive personal information. Please don't paste anything into Wren that you wouldn't want stored to run the product.
We use the information above to run the product and for nothing surprising:
We do not sell your data. We do not use your content to build advertising profiles. We don't share it with third parties except the model provider needed to answer you and standard infrastructure that hosts the app.
This is the important one, so we'll say it plainly: we do not use your chats, Projects, or Frames to train, fine-tune, or improve any AI model. Your conversations are stored so that you can use them — not so that a model can learn from them.
Wren is a product built on top of hosted frontier models; we are not a model lab, and we have no training pipeline that your content feeds. When we improve Wren, we improve the product around the model — the interface, the prompts we write, the Frames tooling — not the model's weights.
Wren does not run its own AI model. To generate a response, your prompt and the relevant context are sent to a hosted model provider through a proxy, which returns the completion that Wren streams back to you. This is a necessary part of how the product works — there is no way to answer you without sending the model something to answer.
We list this data flow and the sub-processor honestly on our Data use notes page. The provider processes the prompt to produce a response; the no-training commitment above covers this path.
Your chats, Projects, and Frames are stored so they persist between sessions. We keep them until you delete them or delete your account. Anonymous visitor sessions and their chats are kept for a limited window and then cleared.
Usage and diagnostic signals are kept only as long as they're useful for running and fixing the app, then aged out. When you delete content, we remove it from active storage; residual copies in routine backups age out on their own cycle and are not used for anything.
Wren uses a small amount of browser storage to function — no advertising trackers, no third-party ad cookies.
You can clear this at any time from your browser. Clearing it signs an anonymous session out and forgets local preferences.
Your data is yours to manage.
Need help with any of this? The Help center walks through it, or contact us.
Wren is intended for adults. You must be 18 or older to use Wren. We don't knowingly collect information from anyone under 18. If you believe a minor has used Wren, contact us and we'll delete the associated data.
If we change how Wren handles data, we'll update this page and move the "Last updated" date. Because this is a reference build, changes here are about keeping the description accurate rather than expanding what we collect.
Questions about privacy, or a request to delete your data? Reach us at privacy@s-d753c7de0a328440.pages.dev, or start in the Help center. Related reading: our Terms of Service and Data use notes.