DCP
Control how AI agents use credentials and private data
Inspect coding-agent data exposure and mask supported sensitive inputs

shim-cli is an open-source local utility for inspecting sensitive information in AI coding workflows. It detects items such as secrets, email addresses, account identifiers, and database connections. Where a supported client's hooks allow rewriting, it can replace detected values with placeholders before the model receives them.
A session watcher reports Claude Code token usage, costs, and detected exposures, while an audit command examines stored session history. Support differs by client: some paths can only be reported or blocked, and some traffic is not inspected. The tool runs on macOS and Linux and is intended to reduce exposure, with best-effort detection rather than a guarantee that all sensitive data is caught.
Tried it? Share a useful thought or ask a question.
Good conversations start with a hello.
Sign in to comment