Skip to content

Releases

v0.2.0

Choose a tag to compare

@github-actions github-actions released this 15 Oct 23:15
  • A menu bar app. It records every agent you start, turns red while one runs, sends a notification for each thing to look at and keeps a window per run with its timeline. Download bodycam-app.zip from the release.
  • bodycam claude, or any other command, records that command in the same terminal. bodycam asks for your password once for Apple's eslogger and runs the command as you. The live feed goes to ~/Library/Logs/bodycam/live.log.
  • bodycam --demo records a harmless pretend agent so you can see what bodycam flags.
  • Catches git push --force when options come before push, like git -C dir push --force.
  • Runs of sh script.sh or node app.js are named after the script, and a program that restarts itself, like /bin/sh, shows up once.
  • The live feed and the summary fit the terminal width.
  • Reports no longer look up hostnames for network addresses, so nothing leaves the Mac.

v0.1.1

Choose a tag to compare

@github-actions github-actions released this 15 Oct 23:14
  • Values in command lines that look like keys or tokens, such as API_KEY=..., Bearer ... or --password=..., are masked in the live feed and the reports.
  • bodycam --version shows the right version when installed with go install.

v0.1.0

Choose a tag to compare

@github-actions github-actions released this 15 Oct 23:14
  • First release. sudo bodycam records every AI coding agent you start: the commands it runs, the files it reads and changes, the background items it registers and the addresses it connects to. When the agent exits you get a summary in the terminal and a full report in ~/Library/Logs/bodycam.
  • Flags reads of secrets such as SSH keys, cloud credentials, token files, .env files, the keychain and browser data, changes to shell startup files, launch agents, SSH and git config, and risky commands like sudo or piping a download into a shell.
  • Tools reading their own credentials, like gh reading its token, are listed but not flagged.
  • --save writes the agent's events to a file that --replay turns back into the same report.