Hey Product Hunt 👋 I build with several Claude Code agents running in parallel — each in its own git worktree, each on its own branch. The tooling for that has gotten good: Claude Squad, Conductor, Vibe Kanban all keep agents from stepping on each other while they work. But nothing helps at the moment the work has to come back together. You find out two branches conflict when you try to merge them, one at a time, after every agent has already finished. Reviewing became my bottleneck, not generating. So I built Braidkeep. Three things: It predicts. It runs git merge-tree across every pair of in-flight branches — read-only, in memory, never touching your working tree — and shows you a live conflict matrix. When an agent commits, the matrix updates and a cell flips. It enforces scope. You declare what files each agent owns in braidkeep.yaml. A Claude Code hook catches out-of-scope edits at write time; a CI check catches them regardless of how the agent was invoked. It sequences. It finds sets of branches with zero pairwise conflicts (merge those now, in any order) and orders the tangled remainder least-conflict-first, so you resolve the hard merges last, with the most context. Two honest caveats, because you'll find them anyway: It predicts mergeability, not correctness. If one agent renames a function and another calls the old name in a different file, git merges it cleanly and Braidkeep says clean. Semantic conflict detection is a much harder problem and I'm not claiming to have solved it. And the merge batches are exact, but the ordering is a greedy heuristic — optimal ordering is NP-hard. It's Apache-2.0 and everything is free. No cloud tier, no signup, single binary. One thing I'll admit: Braidkeep was built by five parallel agents working in separate worktrees, and its own CI runs braidkeep check on every PR — so if any workstream had strayed outside its declared scope, the build would have failed. It didn't. That's the closest thing to a real test I could give it. What's your worst parallel-agent merge story? I want to know which conflict shapes I'm still missing.

Braidkeep

Catch merge conflicts before your AI agents break main

Catch merge conflicts before your AI agents break main