A lot of teams are trying to figure out how to use AI collaboratively and as a team.

I built Draft to help teams with this problem.

It focuses on automatically pulling/pushing context edits to a shared (private) git repo by having Claude Code analyze your agent sessions, Slack threads, and Granola notes/transcripts.

It's human in the loop so you can verify what is being added to context. That context can then be injected into your agents starting with Claude Code, Codex, OpenClaw, and Hermes.

Its free and open source and contains a macOS app, CLI, and skills so agents can work natively with it.

Would love to know how teams are solving this with their teams or any feedback on the project!