A question about one module can sweep in files you weren't looking at: the config with live credentials, the seed script with real user data. You think you're asking about your code. Cursor is assembling context from across the repo.
The control for this sits between your machine and the model. It finds secrets and regulated data in the context an editor gathers, and masks them before they're sent. Locke's local proxy does that for the AI endpoints on its covered list, Claude Code and GitHub Copilot among them. Cursor's own endpoints are not on that list today. Read this page as the argument, not a coverage claim.
The risk with Cursor
Cursor indexes and retrieves across your whole project, so one question can sweep in credentials or real customer data from files you weren't looking at. Locke does not screen Cursor's endpoints today, so it does not currently reduce this risk.
How Locke helps
Locke does not screen Cursor today. Its local proxy screens what a tool on your machine sends to an AI API, and Claude Code and GitHub Copilot are on its covered list. Cursor's own endpoints are not.
Locke is free for one person
It runs on your device and masks sensitive data before a prompt is sent, on the AI tools it covers. Teams starts at $14.99 per seat a month.
