LOCKE FOR CLAUDE
Paste long documents into Claude without pasting your secrets.
Claude's draw is the size of what it can hold: entire contracts, full meeting transcripts, multi-file codebases, hundred-page policy PDFs — all in a single prompt. That capability quietly changes the risk profile. Where ChatGPT use tends toward short snippets, Claude invites you to drop the whole document in, which means a single careless paste can carry far more regulated data than the user realises.
The long-context habit also makes the data harder to eyeball. Nobody re-reads a 40-page transcript checking for the one paragraph with a client's account number before sending it to claude.ai or the Claude desktop app. The bigger the paste, the more likely something sensitive rides along unnoticed.
Locke was built for exactly this: scanning the full prompt, however long, and catching the embedded identifiers a human skim would miss — so you get Claude's reasoning over your whole document without shipping the parts that should never leave.
The risk with Claude
Large context windows encourage whole-document pastes, so a single prompt to Claude can carry far more sensitive data than a quick question — and the volume makes it almost impossible to manually spot the regulated fields buried inside.
How Locke helps
Locke scans the entire prompt on-device — not just the first few lines — and masks identifiers wherever they appear before the document reaches Anthropic. For developers using Claude Code specifically, Canary is a free, open-source plugin that surfaces how much PII a coding session has already exposed.
Keep using Claude — without the exposure
Locke runs entirely on your device. Sensitive data is detected and masked before any prompt is sent, so nothing confidential ever leaves your machine. Pricing for Locke, the desktop app, is coming soon; Canary is free and open source.
