Product launch2 min read
Orca: a development environment for many coding agents at once
Orca by Stably AI runs Claude Code, Codex and other coding agents in parallel – each in its own git worktree. A coordinator agent can split up tasks and merge the results.
Anyone who works with coding agents knows the pattern: one agent works, you wait, review, give the next instruction. Things only speed up when several agents work on different tasks at the same time – and that is exactly what Orca is built for. Its maker, Stably AI, a Y Combinator-backed company in San Francisco, calls it an "agent development environment".
A note on the name: "Orca" is used by several products – including models from Microsoft and a cloud security company. This is about the desktop app from Stably AI.
What Orca does
Orca is a free, open-source desktop app (MIT license) for macOS, Windows and Linux. It bundles chat, terminal, browser, diff view and a CLI for coding agents such as Claude Code, Codex, OpenCode, Gemini or Copilot. You bring the agents yourself – Orca uses your existing subscriptions and accounts.
The core ideas:
- One worktree per agent. Each agent works in its own git worktree with its own files, terminal and session. The agents don't get in each other's way.
- Compare. Give the same task to several agents and look at the results side by side.
- Orchestration. A coordinator agent breaks down a larger task, starts worker agents in separate workspaces and merges the results.
- Runs without your laptop. Agents can run on SSH hosts, an Orca server or a cloud VM per worktree and keep working when your computer is closed.
- Mobile. iOS and Android apps notify you when an agent needs a decision, and you can reply right there.
There is also a built-in browser with a design mode and mobile emulator, GitHub and Linear integrations, scheduled automations and plugins.
Who it's for
Orca targets developers who already work with coding agents and want to take the next step: from a single assistant to a small "fleet". For agentic companies, the orchestration is the most interesting part – it shows what dividing work between agents can look like in practice.
What to watch out for
- Cost. Parallel agents mean parallel use of your subscriptions and API quotas.
- Security. Agents with browser and computer control have more access than a chat assistant. Start in a test repository.
- Pace. The project started in March 2026 and often ships several versions a week. Expect changes.
- Review is still mandatory. More agents produce more code – and more to review. The diff view is no substitute for a real review.