About nerdr
nerdr is a terminal app for Linux and macOS that runs several coding agents at once, each in its own Git worktree. You give each task a tab; nerdr makes a branch and a worktree for it and starts the agent you pick. While the agents work, you see which of them need you, answer their questions, and review what they changed, without leaving the terminal.
Above, a new tab gets Codex and starts on its task, and the status lights show which agents work, wait or are done across three repositories. Claude’s changes are shown in the review panel beside it, the tests run in a terminal, and Codex reviews Claude’s work. The waiting list then takes you to the agent with a question.
Why nerdr
Section titled “Why nerdr”Coding agents are good at working alone for a while. The hard part is everything around them:
- They get in each other’s way. Two agents in the same checkout overwrite each other’s changes. nerdr gives each task its own branch and worktree, so agents work side by side without clashing, and your main checkout stays yours.
- You lose track of them. With five agents going, which one is asking a question? Which is done? Status lights, a waiting list and a short summary above each agent tell you at a glance, and take you to the one that needs you.
- Reading their work takes time. nerdr shows the changes beside the agent, word by word. Ask the agent to explain them, or to walk you through them step by step. Have a second agent review the work and send its findings back for fixes.
- Getting it merged is tedious. Commit, push, open a pull request, send review comments or a failing CI log back to the agent, and close the tab when the pull request is merged, worktree and branch included.
What you get
Section titled “What you get”- Tabs, branches and worktrees: a tab per task, from
main, a GitHub issue, a pull request, another tab’s branch, or several repositories at once. - The workbench: the agent, with terminals, more agents and code reviews beside it in the same worktree.
- Agents working together: agents send each other messages, start more agents, hand work to another agent, and split work into new tabs.
- Code review: a second agent reviews the work and sends its findings back, round after round.
- Progress and notifications: status lights, the waiting list, desktop alerts, rate limits and search through past conversations.
- The review panel: code changes (F1), GitHub (F2), tasks (F3), skills (F4) and plugins (F5), beside the agent.
- Remote machines: run nerdr inside herdr to keep agents going when you disconnect, or on another machine.
Supported agents
Section titled “Supported agents”| Agent | Command |
|---|---|
| Claude Code | claude |
| OpenAI Codex CLI | codex |
| GitHub Copilot CLI | copilot |
| opencode | opencode |
| Cursor CLI | cursor-agent |
| Gemini CLI | gemini |
| Antigravity CLI | agy |
| Grok Build | grok |
nerdr lists the agents it finds on your PATH. Each can have its own arguments, environment and wrapper, such as a sandbox, and you can add agents it does not know. See Agents and sandboxes.
Inspired by herdr
Section titled “Inspired by herdr”nerdr builds on herdr’s status lights, mouse controls and rules for detecting an agent’s state; see NOTICE for attribution. nerdr adds worktrees, task tracking and review. The two work well together: run nerdr inside herdr to keep agents running when you detach, or on another machine.
Contact
Section titled “Contact”Email contact@nerdr.app, or open an issue for bugs and feature requests. For security problems, see the security policy.
License
Section titled “License”nerdr is open source under the Apache License 2.0.