Run five agents. Know which one needs you.
Run many Claude Code agents at once, each in its own git worktree, and see which one needs you: working, needs you, delegating or idle, across every repo.
Everything around
Plexus runs the real Claude Code CLI, one session per task, each in its own git worktree. These pages show the parts of the app around it, in the app itself.
Right now, in the demo
Each bar is as wide as the number of sessions in that state. Click one to open it in the window below.
00.0 / 25 s
Review the agent's work before anyone else sees it.
A PR-style diff of each Claude Code session's worktree, inline comments the agent acts on, AI review, a review-and-fix cycle and the PR conversation.
Every session gets its own checkout, ready to run.
Every Claude Code session gets its own git worktree, your .env files, a block of dev-server ports, and your Create, Run and Teardown hooks at three levels.
Git, pull requests and issues, next to the agents.
A git client per Claude Code session and per repo, GitHub and Bitbucket pull requests with CI status, and GitHub Issues and Projects boards in one window.
Split the change. Merge it in order.
Stack Claude Code sessions on each other's branches. Plexus keeps every branch synced as the agents work, shows the stack as a map, and merges it in order, as a GitHub stack or through a merge queue.
Set it up once for the whole team.
Organizations, shared workspaces, managed settings, a shared prompt library and sync for teams on Claude Code. Seats are paid; the core app stays free.
Weighing it up against
Run your agents
Plexus is free to download for macOS, Windows and Linux. It runs the Claude Code you already have, with your existing login.