Git Worktrees for Multi-Agent Development

If you've tried running multiple AI coding agents on the same codebase simultaneously, you've hit the problem: they step on each other's changes. Agent A modifies a file while Agent B is reading it. Merge conflicts pile up. Builds break for reasons neither agent caused.

The fix is surprisingly simple and already built into git.

How Worktrees Solve It

git worktree add creates a new working directory linked to the same repository but checked out on a different branch. Each agent gets its own directory, its own branch, and its own file state. They share the same git history but can't interfere with each other's uncommitted changes.

git worktree add ../project-agent-1 -b feature/agent-1
git worktree add ../project-agent-2 -b feature/agent-2

Now agent 1 works in ../project-agent-1 and agent 2 works in ../project-agent-2. They can both compile, run tests, and make changes without conflicts.

The Merge Step

The tradeoff is that you still need to merge the branches eventually. But merging two completed feature branches is a known, solvable problem. It's much better than two agents fighting over the same files in real time.

In practice, you want each agent working on a different part of the codebase. Worktrees enforce the branch isolation but they don't prevent logical conflicts if both agents are modifying the same module's behavior.

Why This Beats Other Approaches

You could use Docker containers or separate clones, but both waste disk space by duplicating the entire repo. Worktrees share the .git directory, so they're lightweight. Creating and destroying them is nearly instant.

For anyone running parallel Claude Code sessions or similar multi-agent setups, worktrees are the practical answer to file-level isolation without the overhead of full environment duplication.

Fazm is an open source macOS AI agent. Open source on GitHub.