Orca Puts All Your Coding Agents in One Free App

One App Instead of Five Terminal Windows

Anyone running Claude Code, Codex, or similar CLI agents knows the drill: multiple terminal windows, constant alt-tabbing, and losing track of which agent is stuck waiting for permission. Orca, built by the Stably AI team, replaces that mess with what they call an ADE — agent development environment — rather than an IDE. It's a cockpit for launching, watching, comparing, and merging the output of coding agents.

Orca is free and MIT-licensed. It doesn't proxy your agent traffic through its own servers — you bring your existing subscriptions (Claude, Codex, etc.) and Orca just orchestrates them.

• Install it via Homebrew on Mac, the AUR on Arch, or a direct download from orca.dev for Windows/Linux.
• Bring your own accounts — no extra subscription cost, since it runs your existing Claude, Codex, Grok, or other agent logins.
• Expect broad support — over 30 CLI agents are supported, including Claude Code, Codex, Gemini, Kilo Code, and others, with auto-detection of ones already installed on your machine.

Parallel Work Trees and a Real Dev Environment

The core mechanic is git work trees — real ones, not copies. Every task gets its own isolated work tree, so multiple agents can work on the same repo simultaneously without conflicts. You can even fan one prompt out to several agents and keep whichever result is best.

Each work tree functions as a mini IDE: a Monaco-based code editor (the same engine VS Code uses), a terminal with WebGL rendering, a diff viewer with inline commenting, and attribution tracking that shows which lines were written by AI versus by you.

• Run agents in parallel on the same repo via isolated work trees to compare results or race solutions.
• Review like a senior dev — annotate diffs inline and send comments back to the agent as rework instructions.
• Commit and push directly from Orca, with integration for GitHub, GitLab, Bitbucket, and Azure DevOps.

Built-In Browser, Design Mode, and Orchestration

Each work tree can carry its own embedded browser tab, so a dev server sits right next to the agent building it. An element picker lets you click a UI element to copy its HTML/CSS straight to the agent instead of describing it in words. A screenshot annotation tool sends scribbled markup directly to the agent too. An embedded iPhone simulator and a computer-use mode (with accessibility and screen-recording permissions) let agents interact with mobile or desktop apps directly.

Orca also includes an orchestration skill so agents can hand off tasks to each other — running phased workflows, splitting work across parallel child agents, or breaking large changes into stacked pull requests. An automations tab works like cron for agents, running scheduled repo audits, release readiness checks, and stuck-task alerts.

• Use the element picker to hand agents precise UI targets instead of long descriptions.
• Set up automations for recurring checks like dependency audits or stuck-task detection.
• Delegate between agents using the orchestration skill so one agent can hand off work to another mid-task.

The Payoff

Orca collapses a multi-terminal, multi-tool workflow into one application, saving the constant context-switching and manual git-worktree setup that used to eat into actual coding time. Because it's free, open source, and works with subscriptions you already pay for, there's no added cost to try it — just time saved watching, reviewing, and merging agent work in one place instead of five.