A Multi-Agent IDE to Run Claude Code, Codex, and Others in Parallel Git Worktrees
Orca is a desktop and mobile IDE designed to run multiple AI coding agents—such as Claude Code, Codex, and Grok—concurrently. It leverages Git's worktree mechanism to give each agent an isolated working directory, eliminating the need for stashing or branch juggling. Users can observe and control all agents from a single interface with tabbed panes, built-in diff review, and direct GitHub Issue/PR integration. It's built for developers who rely on CLI-based coding agents and need to handle multiple features or refactors in parallel.
Orca is a desktop application that lets developers run multiple AI coding agents—like Claude Code, Codex, OpenClaude, Grok, and dozens of others—side by side in isolated Git worktrees. Think of it as an IDE for agent workflows: each agent gets its own clean workspace, and you can switch between them, review diffs, and commit changes without leaving the app. It runs on macOS, Windows, and Linux.
Orca 是一个桌面应用，让开发者可以并排运行多个 AI 编程 agent（如 Claude Code、Codex、OpenClaude、Grok 等数十种），每个 agent 独立在自己的 Git worktree 中运行。你可以把它理解成一个针对 agent 工作流设计的 IDE：每个 agent 拥有一个干净的工作区，无需切出应用即可在不同 agent 间切换、审查差异并提交更改。支持 macOS、Windows 和 Linux。
When developers work with multiple AI agents, each agent typically operates in its own directory or branch. This leads to constant stash/pop cycles, branch switching conflicts, and a fragmented view of what each agent is doing. Orca solves this by leveraging Git worktrees—each agent gets a physically separate checkout of the same repository, so they never interfere. You can launch two agents on the same repo simultaneously, compare their outputs, and pick the best approach.
当开发者同时使用多个 AI agent 时，每个 agent 通常运行在自己的目录或分支中。这会导致频繁的 stash/pop、分支切换冲突，以及各 agent 工作状态的碎片化。Orca 利用 Git worktree 解决这个问题——每个 agent 获得同一个仓库的物理独立检出，互不干扰。你可以在同一个仓库上同时启动两个 agent，比较它们的输出，选择最佳方案。
Orca supports over 30 AI coding agents out of the box—from Claude Code, Codex, and Gemini to niche ones like Kilocode, Qwen Code, or Charm. It doesn't require a login; you bring your own subscriptions for the agents you use. Each agent runs as a terminal process inside the Orca UI, which shows live status tabs. You can see which agents are active, view their output in split panes, and switch between worktrees with a click.
Orca 内置支持 30 多种 AI 编程 agent，从 Claude Code、Codex、Gemini 到小众的 Kilocode、Qwen Code、Charm 等。它无需登录，你直接使用自己的 agent 订阅即可。每个 agent 在 Orca 界面中以终端进程运行，显示实时状态标签页。你可以一目了然地看到哪些 agent 正在活跃，在分屏面板中查看输出，点击即可切换 worktree。
Beyond just running agents, Orca provides built-in source control: you can review AI-generated diffs inline, make quick edits, and commit directly from the app. Each worktree is automatically linked to GitHub pull requests, issues, and Actions checks. SSH support lets you run agents on remote machines. Notifications alert you when an agent finishes or needs your attention.
除了运行 agent，Orca 还内置了源码控制功能：你可直接内联审查 AI 生成的差异、快速编辑并提交。每个 worktree 自动关联 GitHub 的 PR、issue 和 Actions 检查。SSH 支持让你在远程机器上运行 agent。当 agent 完成任务或需要你的注意时，会收到通知提醒。
Orca can be downloaded from onOrca.dev or via GitHub Releases. macOS users can install with Homebrew: brew install --cask stablyai/orca/orca. Arch Linux users can install from AUR: yay -S stably-orca-bin or yay -S stably-orca-git. Once installed, launch the app, choose a repository, and pick which agents to run. Each agent spawns in its own worktree. A mobile companion app (iOS and Android) lets you monitor agents from your phone.
可以从 onOrca.dev 或 GitHub Releases 页面下载 Orca。macOS 用户可通过 Homebrew 安装：brew install --cask stablyai/orca/orca。Arch Linux 用户可以从 AUR 安装：yay -S stably-orca-bin 或 yay -S stably-orca-git。安装后启动应用，选择仓库，再选择要运行的 agent。每个 agent 在自己的 worktree 中启动。移动端 companion app（iOS 和 Android）可让你在手机上监控 agent。
Orca is ideal for developers who regularly experiment with multiple AI agents on the same codebase—trying different prompts, models, or approaches in parallel. It's also useful for teams that want a unified dashboard for agent-based code generation. Caveats: Orca is a GUI application; it's not a headless CLI tool. You need to have the agent CLI tools (e.g., Claude Code CLI) installed and authenticated separately. The worktree approach works best for individual repos; cross-repo orchestration is not its primary focus.
Orca 适合经常在同一代码库上用多个 AI agent 做实验的开发者——并行尝试不同的提示、模型或方法。对于希望有一个统一面板管理 agent 代码生成的团队也很有用。注意点：Orca 是 GUI 应用，不是纯 CLI 工具。你需要预先单独安装并认证你使用的 agent CLI 工具（如 Claude Code CLI）。worktree 方式最适合单一仓库；跨仓库编排不是其主要关注点。