Running more than one AI coding agent at the same time usually means either cloning the same repo multiple times or watching agents overwrite each other's work. Modulus, a new macOS desktop app, gets around that using git worktrees — Git's built-in mechanism for maintaining multiple working directories off a single repository — so each agent operates in its own isolated environment without the overhead of duplicate clones or separate IDE windows.
The app is free and currently supports Claude Code. Developers can run agents on separate tasks concurrently — patching a production bug in one worktree, building a new feature in another — without the two interfering. When the agents are done, everything lands in a single review interface where developers can inspect the changes and ship them as pull requests.
The more substantive piece is what Modulus calls Project Memory: a shared knowledge layer that automatically supplies every running agent with the same current context about the codebase — API schemas, dependency trees, recent commits. The alternative is copying that context manually into each new agent session, which works fine until you have three agents running and the project has changed since you last pasted anything. Modulus keeps it consistent across sessions and across repositories without developer intervention.
Linux and Windows versions are on a waitlist, which suggests the team is planning for a broader release rather than staying Mac-only.
Modulus sits in a category that's still taking shape: tooling that coordinates agents rather than acting as one. Whether that layer ends up as a standalone product or gets folded into the agents and editors it wraps is an open question. For now, Modulus is building for the developers who are already running multiple agents and spending too much time managing the plumbing.