Blockodile

What Blockodile does

Blockodile gives a team one policy for how its coding agents may work, and enforces it on each developer's machine. The blockodile CLI puts a blocking hook in front of every agent action, scans for secrets, and writes the agent's own managed settings where they exist. Every ask, warn or block lands in an audit log with its snippet redacted. Owners run it from a web console (teams, invitations, the policy, members and the log), and the CLI signs in with a device code. Claude Code is supported today, with Codex, Gemini CLI and Cursor next, then a browser extension. Beside it sits Autopilot: it picks up tickets from Jira, Linear or GitHub and the reviews people ask for, works each one in its own headless Claude Code session and worktree, opens the pull request and asks for reviewers. Ticket text is screened before a run sees it, and runs never merge or deploy. A native desktop app, written in Rust on GPUI, shows each run as a terminal tab and lives in the macOS menu bar with Run now, today's cost and the log.

Features

• ·One policy per team, enforced by a blocking hook before every agent action
• ·Secret scanning, with snippets redacted in the audit log
• ·Web console for teams, invitations, the policy, members and the audit log
• ·Native desktop app in Rust on GPUI, in the macOS menu bar, showing each run as a terminal tab
• ·Autopilot: tickets from Jira, Linear or GitHub become pull requests from headless Claude Code runs
• ·Ticket text is screened before a run sees it, and runs never merge or deploy