Avoid merge conflicts across git worktrees for parallel AI coding agents
Early conflict detection via merge simulation—directly fixes Claude Code's worktree blindness problem.
Teams running multiple AI coding agents in parallel (Claude Code, Cursor, Codex) on shared codebases
What's still unsolved is conflicts across active worktrees, which often surfaces late.
I built Clash (OSS, Rust) to detect potential conflicts earlier using read-only merge simulation across worktrees during edits.
- clash check <file> for pre-edit checks - clash status for a cross-worktree conflict matrix - clash watch for live monitoring - Claude plugin hook support for Write|Edit|MultiEdit
Quick setup (CLI + Claude plugin):curl -fsSL https://clash.sh/install.sh | sh && \ (claude plugin marketplace add clash-sh/clash || true) && \ (claude plugin install clash@clash-sh || claude plugin update clash@clash-sh)
Would love feedback or your thoughts on this.Git worktree multiplexing lets one agent write tests while another refactors—750+ merged PRs prove the workflow.
Git worktrees per agent prevent merge conflicts, but audience remains narrow without multi-agent proof.
Runs ten Claude agents in parallel with global hooks tracking status across worktrees.
Git worktree isolation for parallel agent sessions solves the context-loss problem elegantly.
Worktree pooling for parallel Claude sessions, but only solves friction if using both Git worktrees and AI agents.
Git worktree pain solver that actually ships—fleet mode runs 5 agents in parallel.