Lich: The Developer Tool That Lets AI Coding Agents Work in Parallel Without Stepping on Each Other

The Parallel Agent Problem

If you've been experimenting with AI coding agents lately, you've probably run into this wall: running one agent is powerful, but trying to scale that to multiple agents working in parallel? That's where things get messy fast.

Each agent needs its own development environment. Each one wants to install dependencies, modify files, and run tests. Stack them on top of each other and you've got dependency conflicts, file contention, and a headache-inducing debugging session.

RPate97's Lich project is an elegant solution to this problem—and it leverages something already built into Git.

Worktrees: Git's Best-Kept Secret

Most developers use Git the same way: one working directory, one branch at a time. But Git worktrees have been lurking in your toolbelt for years, quietly offering something powerful: multiple working directories from a single repository, each checked out to a different branch.

Lich takes this capability and supercharges it. Instead of just checking out different branches, Lich creates fully isolated development stacks—complete environments where your coding agents can operate independently without bumping into each other.

Why This Matters for AI Development

When you're orchestrating multiple coding agents—whether for parallel feature development, testing different approaches, or running independent review workflows—you need guarantees of isolation. Lich provides that guarantee at the filesystem level, the environment level, and the process level.

Think about what becomes possible:

• Agent A can refactor the authentication module while Agent B builds the new dashboard feature
• Multiple agents can run tests simultaneously without fighting over port assignments
• You can experiment with different approaches in parallel and merge only what works

The Developer Experience

What makes Lich particularly interesting is the simplicity of the premise. Rather than spinning up Docker containers or managing complex Kubernetes orchestrations, you're working with native Git behavior—just amplified.

For teams building agent-based development pipelines, this could be the missing piece that makes multi-agent architectures actually practical at the development stage rather than just in theory.

The project is still relatively new, but the concept demonstrates exactly the kind of infrastructure thinking we need as AI-assisted development moves from novelty to standard practice.

If you're building anything involving multiple coding agents or parallel development workflows, Lich is worth keeping on your radar. It's a reminder that sometimes the most powerful solutions build on tools we already have—we just needed someone to show us what they could become.

Check it out: RPate97/lich on GitHub