Your manuscript stays yours.

Data handling

What stays local and what travels.

• Draft text, marks, dictionary entries, scan state, and local settings stay in your browser unless you ask for a Read or a Write.
• If you request access or use account features, KRW may store your email address and account identifiers needed to operate that flow.
• Your manuscript is never used to train a model. When you request a reading, use Ask-AI, or apply a rewrite, the relevant passage and context pass through KRW's proxy to the hosted provider processing your request, then return a response. Nothing is stored, copied, or reused afterward. BYOK requests take the same path: your key rides the request through KRW's proxy to your chosen provider, is used for that request, and is never stored.
• Hosted free and Pro run on KRW's provider accounts. Bring-your-own-key works on every tier and keeps billing on your side.
• If you connect Google Drive, available in the Reader, Weaver, and Rewriter, your manuscript saves to your own Drive as a file you own, never to a KRW server. Each app connects separately and keeps its own backups; one app cannot reach another app's files. KRW asks Google for your account email, so it can show you which account is connected, and for access only to the files it creates for you, so it cannot see the rest of your Drive. To save a version or bring one back, that file passes through KRW's proxy in the open and is not stored there afterward. When automatic sync is on, a Pro option, the file also passes through the proxy on a periodic background schedule, and nothing is stored there afterward either. You can disconnect Drive at any time, and removing KRW's access from your Google account stops it reaching those files. Deleting a manuscript in an app leaves its Drive backup untouched: the backup stays your file, listed for restore whenever you want it back, and you remove it from your own Drive whenever you choose.
• This site uses Cloudflare Web Analytics, which is cookieless. Cloudflare receives the visited URL, the referrer, browser type, and the IP address that delivered the request, and uses that data to count visits and screen out bot traffic. No cross-site tracking, no profile building.

Request path

How the request travels.

1. Your browser

   Your draft and revision memory live here first. Nothing leaves until you ask for a Read or a Write, or until Drive backup runs as described above.

2. KRW proxy

   KRW passes the request to the model provider and hands the response back. The proxy carries the passage through; it stores nothing.

3. Model provider

   Hosted routes run on paid APIs that do not train on inputs by default. Google's paid Gemini API and Anthropic's commercial API both exclude API traffic from training. Their current documentation is the source of truth.

Provider docs

The provider's current documentation is the source of truth.

Policies and models change. If the work is sensitive, read the provider docs before you send it through hosted AI.

• Hosted free uses Google Gemini on Google's paid API. Read the Gemini model docs and API terms.
• Pro uses Claude via Anthropic's commercial API. Read the Claude model docs and Anthropic privacy docs.
• Bring your own key works on every tier. Your provider bills you directly. Your key lives in your browser, rides only your own requests through KRW's proxy, and does not burn hosted usage. What the provider does with the text is governed by the account and terms on your key.