Orca is a free, open-source agent development environment (ADE) for macOS, Windows, and Linux. You run coding agents like Claude Code, Codex, Cursor CLI, and OpenCode in parallel. Each task gets its own git worktree, terminal, and browser, so agents never touch each other's work. Review diffs, annotate lines for the agent, and open PRs from one app. It suits developers who use several CLI agents and want to compare and merge the best result.

• Parallel Git Worktrees: Every task runs in its own isolated git worktree. Send one prompt to several agents, compare what they produce, and merge the best version.
• Broad Agent Support: Claude Code, Codex, Cursor CLI, Gemini, Copilot, OpenCode, and Pi come preconfigured. You can also run Grok, Amp, Goose, Cline, Kiro, Qwen Code, or any other CLI agent.
• Inline Diff Annotations: Add markdown comments to specific diff lines, batch them, and send them back to the agent. You can also check CI, resolve conflicts, and create PRs without leaving the app.
• Design Mode Browser: Each worktree gets an embedded Chromium browser. Click any UI element and its HTML, CSS, and a cropped screenshot go straight to your agent.
• GPU-Rendered Terminals: WebGL terminals with unlimited splits, scrollback search, and scrollback that comes back after a restart.
• SSH Worktrees and Remote Servers: Run agents on a remote machine with full file editing, git, and terminals. Auto-reconnect, port forwarding, and self-hosted Orca servers are included.
• GitHub, Linear, and Jira Integration: Browse PRs, issues, and boards, review and approve PRs, and spin up a worktree straight from a task.
• Orca CLI and Automation: Agents can drive Orca themselves to create worktrees, take checkpoints, run scheduled automations, and use the computer. A skills registry with MCP is also available.

• Free and open source: Orca is MIT-licensed with no paid tier for the desktop app, so you can try it on real projects at no cost.
• No more agents colliding: Isolated worktrees let several agents work on the same repo at once without overwriting each other's changes.
• Works with the agents you already pay for: You bring your own Claude, Codex, or other subscriptions, and you aren't locked into one model vendor.
• Cross-platform support: Parallel worktrees work on macOS, Windows, and Linux. Some agent wrappers only run on macOS.
• Check in from your phone: The iOS and Android companion shows live agent status and usage, and it lets you switch accounts while you're away from your desk.
• Usage tracking built in: Claude and Codex usage and rate-limit resets appear in the app. You can also hot-swap Codex accounts without logging in again.

• No models included: Orca doesn't provide AI models itself. You need your own Claude, Codex, or other agent subscriptions, so your real cost depends on those plans.
• Orchestration, not inline completion: Orca focuses on running and supervising CLI agents. If you mainly want tab autocomplete and inline chat in one editor, a traditional AI code editor may suit you better.
• Runs on your own hardware: The app runs on your desktop by default. It isn't a hosted VPS, so remote or always-on work needs machines or cloud accounts you set up yourself.
• Mobile app has limitations: The iOS and Android companion apps are in beta and rely on a paired desktop or remote Orca server. They support monitoring agents and handling small tasks but are not full standalone development environments.
• Enterprise pricing not published: An Enterprise offering exists, but no plans or prices are listed, so teams need to contact the company for details.

• Racing agents on one task: Send the same bug fix prompt to Claude Code, Codex, and Cursor CLI in separate worktrees. Compare the three diffs side by side and merge the cleanest solution.
• Fixing UI bugs visually: Open your app in the per-worktree browser and turn on Design Mode. Click the broken element, and the agent gets its markup, styles, and a screenshot, so it can fix the problem without long descriptions from you.
• Reviewing AI-written code line by line: Annotate specific lines in the diff viewer with feedback and send the batch back to the agent. You keep the same review standards you use for human code before committing or opening a PR.
• Running agents on a remote box: Connect to a VPS or Mac mini over SSH and run agents there with full git and terminal access. Check progress from the mobile companion while you're away.
• Working through a ticket backlog: Open worktrees straight from GitHub issues, Linear, or Jira items. Several agents can then work through tickets in parallel while you review and ship.
• Standardizing agent use across a team: Engineering leads can roll out Orca with approved providers and organization defaults. Code stays on machines you control, and every change lands as a reviewable PR.