Thadam
Records what AI coding tools do on your Mac (commands, file changes, network connections and prompts) in a tamper-evident local audit log, with a dashboard to review and export it. Nothing leaves your machine.
AI coding agents run commands, edit files and reach the network on your behalf, often faster than you can watch. Thadam is a free, open-source audit trail for that work on macOS.
It watches only the AI tools you name (matched by code-signing identity, not file name) and records what they did: commands run, files read and changed, network connections, prompts, and commands you declined. Rows are grouped into sessions and kept in a local, append-only, hash-chained store, so later tampering shows. Secrets and personal data found in commands are redacted before anything is written, and file contents are never stored.
A local dashboard shows sessions, a Flagged queue (secrets, sensitive paths, writes outside the project, unknown hosts, changes to a toolβs own configuration), search, export, and a coverage timeline that shows when Thadam wasnβt watching.
Everything stays on your Mac. It ships as a signed, notarized installer and is licensed under Apache-2.0. Built in Rust, .NET and React.