If you've got both teammates and coding agents working on the same repo, where does the backlog actually live? Jira/Linear that the agents can't see? A markdown file? Nowhere? Wondering if anyone's found a setup where humans and agents work the same list without it falling apart.
Cairn is a cross-platform desktop app for tracking tasks as Markdown files in your repo. No database. It serves the same tasks to your AI agents over MCP and to you over a clean UI, so an agent and a human always see the same state.
The difference: agents can't fake "done." Every task carries checks like go test ./.... cairn runs them on close and refuses to close if any fail. Git owns the history, every write is signed, and agents hand off for review. Open source.
