Locke works with the AI you already use.

A chat box, an IDE and a local model expose data in different ways. Each page names the exposure, what Locke covers and what it doesn't.

Web & desktop assistant
Locke for ChatGPT
The office scratchpad. Mask the sensitive fields before a prompt reaches OpenAI.
See how it works
Web & desktop assistant
Locke for Claude
Bigger context, bigger pastes. Mask what's sensitive before the document leaves your device.
See how it works
Workspace-integrated assistant
Locke for Gemini
Wired into Gmail, Docs and Drive. Mask what's sensitive before it reaches the model.
See how it works
AI coding assistant
Locke for GitHub Copilot
It reads your open files for context, including the ones with keys and customer data.
See how it works
AI-native code editor
Locke for Cursor
Indexes your whole project, secrets included. Not on Locke's covered list today.
See how it works
Local model runner
Locke for Ollama
No vendor sees the prompt. Your own logs still do.
See how it works
AI answer engine
Locke for Perplexity
Your question goes to a model, then out to the open web.
See how it works
Enterprise productivity assistant
Locke for Microsoft Copilot
Screens the Copilot desktop app. Browser chat isn't screened, and embedded apps have their own coverage limits.
See how it works

Locke, the Sonomos desktop app, and its browser extension screen the AI tools Locke covers on your machine. Download Locke or compare approaches.