Orca: An Open-Source Cockpit for Running Multiple AI Coding Agents at Once

If you use an AI coding agent — Claude Code, Codex, Cursor CLI, whatever — you’ve probably hit the same wall: one agent, one terminal, one task at a time, and a lot of tab-switching once you’re running more than one thing. Orca (GitHub, onOrca.dev) is a desktop app built specifically to fix that. I’ve been using it, and it’s genuinely useful enough to write up.

What it actually is

Orca calls itself an “ADE” — an Agentic Development Environment. In practice it’s a desktop app (macOS, Windows, Linux) that sits on top of whatever coding agent you already use, and gives you a proper interface for running several of them at once instead of juggling terminal tabs.

It’s agent-agnostic by design: Claude Code, Codex, Cursor CLI, GitHub Copilot CLI, OpenCode, Amp, Devin, Goose, and a long list of others are all supported out of the box, plus “any CLI agent” in general — if it runs in a terminal, Orca can drive it.

The core idea: parallel worktrees

The feature that sold me is parallel worktrees: you fan one prompt out to several agents, each running in its own isolated git worktree, then compare the results side by side and merge whichever one actually did the best job. Instead of running an agent, judging the output, and re-prompting when it’s not quite right, you get several attempts at once and pick the winner.

This maps onto something a lot of people doing serious AI-assisted development are already doing manually — running the same task with two or three different agents or prompts to see which one lands — Orca just makes that a first-class, one-click workflow instead of a pile of manual terminal tabs and copy-pasted diffs.

What else is in the box

Terminal splits — Ghostty-class terminal rendering with infinite splits and scrollback that survives restarts.
Design Mode — click any element in an embedded Chromium window and its HTML, CSS, and a cropped screenshot go straight into the agent’s prompt. Useful for “make this button look like that” without describing it in words.
GitHub & Linear, native — browse PRs, issues, and boards inside the app, and open a worktree directly from a task without switching context.
SSH worktrees — run agents on a remote box with full file editing, git, and terminal access, with auto-reconnect and port forwarding.
Annotate AI diffs — comment directly on a diff line and send it back to the agent instead of writing a fresh prompt explaining what to fix.
Mobile companion (iOS/Android) — get notified when an agent finishes or needs input, and send follow-ups from your phone. Handy if you kick off a longer task and don’t want to sit and watch a terminal.
Orca CLI — agents can drive Orca itself, scripting worktree creation, snapshots, clicks, and form fills.

Pricing and how it’s actually built

Orca is free and open source under the MIT license. There’s no markup layer or bundled model access — you bring your own subscriptions (Claude, Codex, whichever agents you use) and Orca just orchestrates them. That’s a meaningfully different model from most “AI IDE” products, which usually resell model access at a margin.

How new (and how active) this project actually is

Worth being upfront about: Orca is young. Its first stable release was in March 2026, and as of this writing it’s already past 24,000 GitHub stars and 1,700+ forks, with releases shipping almost daily (multiple release candidates a week is normal for this project). That pace is a real strength — bugs get fixed fast — but it also means the ground can shift under you between versions. If you build a workflow around a specific feature, check the changelog before updating rather than assuming everything is stable.

Who this is actually for

You’re running more than one coding agent, or the same agent on more than one task, and you’re tired of terminal-tab juggling — this is squarely built for you.
You work across desktop and want to check on long-running agent tasks from your phone without SSHing in.
You want git-worktree-based parallel attempts (run the same prompt three ways, keep the best one) without hand-rolling that setup yourself.

If you only ever run a single agent sequentially on a single repo, Orca is more surface area than you need — a plain terminal is fine.

Getting it

Desktop: onorca.dev/download, or via Homebrew (brew install --cask stablyai/orca/orca) on macOS, or the AUR on Arch Linux.
Mobile: iOS App Store / TestFlight, and an Android APK on GitHub.
Source and docs: github.com/stablyai/orca

Agent orchestration — running several AI agents productively instead of one at a time — is a topic we’re planning to cover a lot more here as we build out deeper, hands-on content. Subscribe to the newsletter if you want that when it lands.