An open-source desktop app by the stablyai team (
stablyai/orca).
What it does
Using coding agents one at a time makes it hard to try several approaches to the same task or work on multiple issues at once. Juggling terminal tabs and hand-managing git worktrees gets tedious fast.
Orca is a worktree IDE built for AI coding agents rather than human developers (an ADE, Agent Development Environment). It runs multiple agents in isolated git worktrees at the same time, then lets you compare each result side by side and merge the ones you want. Run Claude Code, Codex, Grok, and others together on one screen with your own subscription. From the repo: “Orca is the ADE for working with a fleet of parallel agents.”
What you can do
The use cases shown on the official site:
- One task, many agents in parallel — hand the same task to several agents at once and compare the per-worktree results side by side
- Regression testing and validation — several agents validate a flow like checkout in parallel (the site’s “5 agents running parallel validation” example)
- Code review and PR management — handle PRs and issues in-app via GitHub and Linear integration, with inline markdown annotations on diffs
- Remote execution — run agents on remote machines with SSH worktrees and monitor progress from the mobile app
The flow: distribute a task to several agents → each runs in its own isolated worktree → compare the results side by side → pick and merge the one you want.
Key features
-
Parallel worktrees
Runs each agent in an isolated git worktree at the same time, so multiple approaches run side by side without interfering with each other.
-
Multi-agent support
Works with Claude Code, Codex, Grok, Cursor, Gemini, and 30+ CLI agents. Each signs in with its own subscription, with account switching and usage tracking.
-
Built-in terminal and editor
A WebGL-rendered terminal (infinite splits, full scrollback search) and a VS Code-based editor (autosave, drag-and-drop) are built into the app.
-
GitHub and Linear integration
Handle PRs and issues in-app, and leave inline markdown annotations on diffs.
-
Design Mode and remote execution
Design Mode with embedded Chromium to click UI elements and inspect HTML/CSS, SSH worktrees for remote execution, and iOS/Android companion apps.
Usage
Install — onorca.dev/download
| Platform | Build |
|---|---|
| macOS Apple Silicon | DMG (arm64) |
| macOS Intel | DMG (x64) |
| Windows | Windows 10/11 x64 installer |
| Linux | Universal AppImage |
On macOS you can also install the Homebrew Cask.
brew install --cask stablyai/orca/orcaArch Linux has AUR packages — yay -S stably-orca-bin, or stably-orca-git to build from source.
The mobile companion apps ship through the App Store and TestFlight on iOS, and as an APK from GitHub Releases on Android.
Run — open the app, sign in to the agents you want to use (Claude Code and others) with their own accounts, then distribute tasks to run in parallel. CLI agents like Claude Code must be installed beforehand.
Notes
- Who makes it — Orca is built by Stably AI, a startup whose main product is an AI software testing platform.
- Agents installed and billed separately — Orca wraps CLI agents, so tools like Claude Code must be installed first, and their usage and subscriptions are billed to your own accounts.
- Desktop app — runs on macOS, Windows, and Linux, with iOS and Android companion apps for remote monitoring.
- MIT License — published as open source.