DEV Community

Deduped
Deduped

Posted on

Run Claude Code and Codex in parallel git worktrees and track usage with Orca

#ai #claudecode #git #productivity

Running one Claude Code or Codex session is easy. Running five at once on the same repo is where it falls apart: agents overwrite each other's files, you lose track of which terminal finished, and your usage limit vanishes without warning.

stablyai/orca is a desktop app that treats each agent as a first-class worktree. Per its README, you run Codex, Claude Code, OpenCode or Pi side by side, each in its own git worktree, tracked in one place. This post covers the parts that help most with Claude Code and Codex, especially watching usage and swapping accounts.

What it gives you

From the README feature list:

  • Parallel worktrees. Fan one prompt across five agents, each in an isolated git worktree, compare the results and merge the winner. Because each agent has its own checkout, they cannot clobber each other.
  • Account switcher and usage tracking. See Claude and Codex usage and rate-limit resets, and hot-swap accounts without logging in again.
  • Notifications and unread state. Know when an agent finishes or needs attention, and mark threads unread to come back later.
  • Annotate AI diffs. Drop comments on any diff line and send them back to the agent.
  • Terminal splits. Ghostty-class terminals with WebGL rendering, and scrollback that survives restarts.
  • Design mode. Click any element in a real Chromium window to send its HTML, CSS and a cropped screenshot into the agent's prompt.
  • GitHub and Linear. Browse PRs, issues and boards in the app, and open a worktree from any task.
  • SSH worktrees. Run agents on a remote machine with file editing, git, terminals and port forwarding.
  • Mobile companion. Get notified from your phone when an agent finishes and send follow-ups. The README lists iOS and Android apps.
  • Orca CLI. Agents can drive Orca too, with commands such as orca worktree create, snapshot, click and fill.

It works with any CLI agent, so if it runs in a terminal, it runs in Orca.

Get started

Download the app from the project's site or the GitHub releases page; the README points to a Download Orca button and says the team ships daily, so the changelog is the real feature list. Then:

  1. Open a git repository in Orca.
  2. Create a worktree for the task. Each worktree is its own branch and folder.
  3. Launch Claude Code or Codex in that worktree's terminal.
  4. Repeat for as many parallel attempts as you want, for example one worktree per approach to the same bug.
  5. Compare the diffs, comment on lines you want changed, and merge the one you like.

Watching usage across accounts

If you hit rate limits on Claude Code or Codex, the usage tracking view is the feature to try first. The README says it shows usage and reset times for both, and lets you hot-swap accounts without a new login. Use only accounts you own and stay inside each provider's terms.

Caveats

  • Several agents in parallel multiply token use. Watch the usage view.
  • Worktrees share your git history, so disk use is small, but dependency installs (node_modules and the like) happen per worktree.
  • This walkthrough follows the README and we have not run it end to end.

The step-by-step version is on our use-case page: watch Claude and Codex usage and swap accounts with Orca. If you want one app for several agents without worktrees, see control Claude Code and Codex from one app with T3 Code.

Want the setup commands for 50 AI repos people actually ship with? Get the free AI Repos That Actually Ship pack: https://dedupedai.com/ship-pack/?utm_source=devto&utm_medium=article&utm_campaign=shippack

Top comments (0)