Orca — The Next-Gen IDE for Parallel Agentic Development
文章目录
- Worktree-Native Architecture: Every AI agent task gets its own Git worktree. No more branch juggling or stashing — switch between parallel workstreams instantly. Agents stay isolated but tracked in one unified sidebar. Built-In Source Control: Review AI-generated diffs, make quick edits, and commit without leaving the IDE. GitHub integration links PRs, issues, and Actions checks to each worktree automatically. Per Worktree Browser & Design Mode: A built-in browser lets you preview your app while building. Switch to Design Mode to click any UI element and drop it directly into the AI chat as context — no screenshots, no copy-pasting selectors. Multi-Provider Support: Claude Code, Codex, Gemini CLI, OpenCode, Goose, Cursor, and 15+ more agents — all supported. Hot-swap between multiple Codex accounts instantly for the best token economics.
- The Orca community is actively shaping the roadmap. Here are three discussions from GitHub Issues that showcase what's happening:
- Developer @mralexsaavedra opened a PR to add Gemini and OpenCode usage tracking with per-provider Promise.allSettled isolation. The discussion evolved into a deep architectural review: "Hey @nwparker, thanks for the feedback! I've just performed a deep cleanup of the PR. The Linear changes you saw were actually noise introduced during an AI-driven adversarial review (Judgment Day protocol), where the agent suggested optimizations that, while technically sound, were clearly out of scope for this feature." — @mralexsaavedra Reviewer @nwparker pushed back on scope creep while praising the core design: "The React Flight parser regression tests, the cookie filtering, and the per-provider Promise.allSettled isolation are all solid." The final merge required a critical regex fix for workspace-ID parsing — a reminder that even sharp developers make typo-level mistakes under AI review pressure.
- Developer @sasha-id proposed storing worktrees inside the repo itself under a .worktrees/ directory. The feature gained momentum quickly: "This is a good feature. Will try to help land it tomorrow." — @AmethystLiang But reviewer @brennanb2025 surfaced a subtle issue: with .worktrees/ in .gitignore, ripgrep-based tools return zero results for strings that exist in sibling worktrees while Glob still returns files — agents encounter "logically impossible state." The discussion resulted in a cleaner replacement PR (#1129) that addressed the architectural concerns.
- Multiple users requested SSH support to run agents on remote machines. The Orca team confirmed it's actively being developed: "It's coming! This is being actively worked on." — @AmethystLiang Developer @Jinwoo-H acknowledged a related bug: remote repos disappear from the sidebar after restart and need to be re-added. They committed to fixing it once SSH leaves beta. User @TimothyW553 also requested an adjustable SSH timeout beyond the current 15-second default — the team agreed to make it configurable.
- Orca fills a real gap in the AI coding tooling landscape: the ability to run, track, and coordinate multiple agents across repos without chaos. Its worktree-native approach is elegant, its multi-agent support is comprehensive, and the active community is pushing features like SSH and in-repo worktrees forward fast. If you're running more than one AI agent on a regular basis, Orca is worth a serious look. Available for macOS, Windows, and Linux. No login required — just bring your own subscriptions. 🔗 stablyai/orca on GitHub
Orca is an open-source IDE built for developers who want to run multiple AI coding agents simultaneously across different repositories. Instead of juggling tabs or stashing changes, each agent gets its own Git worktree — so Claude Code, Codex, Gemini, and OpenCode can all run in parallel without interference. It's the orchestration layer the multi-agent era has been waiting for.
Created by Stably.ai, Orca started in March 2026 and has already attracted significant community attention. With native support for over 20 different AI agents and a design-first approach to developer experience, it's quickly becoming the go-to tool for power users managing complex, multi-repo projects.
- Worktree-Native Architecture: Every AI agent task gets its own Git worktree. No more branch juggling or stashing — switch between parallel workstreams instantly. Agents stay isolated but tracked in one unified sidebar.
- Built-In Source Control: Review AI-generated diffs, make quick edits, and commit without leaving the IDE. GitHub integration links PRs, issues, and Actions checks to each worktree automatically.
- Per Worktree Browser & Design Mode: A built-in browser lets you preview your app while building. Switch to Design Mode to click any UI element and drop it directly into the AI chat as context — no screenshots, no copy-pasting selectors.
- Multi-Provider Support: Claude Code, Codex, Gemini CLI, OpenCode, Goose, Cursor, and 15+ more agents — all supported. Hot-swap between multiple Codex accounts instantly for the best token economics.
The Orca community is actively shaping the roadmap. Here are three discussions from GitHub Issues that showcase what's happening:
Developer @mralexsaavedra opened a PR to add Gemini and OpenCode usage tracking with per-provider Promise.allSettled isolation. The discussion evolved into a deep architectural review:
"Hey @nwparker, thanks for the feedback! I've just performed a deep cleanup of the PR. The Linear changes you saw were actually noise introduced during an AI-driven adversarial review (Judgment Day protocol), where the agent suggested optimizations that, while technically sound, were clearly out of scope for this feature." — @mralexsaavedra
Reviewer @nwparker pushed back on scope creep while praising the core design: "The React Flight parser regression tests, the cookie filtering, and the per-provider Promise.allSettled isolation are all solid." The final merge required a critical regex fix for workspace-ID parsing — a reminder that even sharp developers make typo-level mistakes under AI review pressure.
Developer @sasha-id proposed storing worktrees inside the repo itself under a .worktrees/ directory. The feature gained momentum quickly:
"This is a good feature. Will try to help land it tomorrow." — @AmethystLiang
But reviewer @brennanb2025 surfaced a subtle issue: with .worktrees/ in .gitignore, ripgrep-based tools return zero results for strings that exist in sibling worktrees while Glob still returns files — agents encounter "logically impossible state." The discussion resulted in a cleaner replacement PR (#1129) that addressed the architectural concerns.
Multiple users requested SSH support to run agents on remote machines. The Orca team confirmed it's actively being developed:
"It's coming! This is being actively worked on." — @AmethystLiang
Developer @Jinwoo-H acknowledged a related bug: remote repos disappear from the sidebar after restart and need to be re-added. They committed to fixing it once SSH leaves beta. User @TimothyW553 also requested an adjustable SSH timeout beyond the current 15-second default — the team agreed to make it configurable.
Orca fills a real gap in the AI coding tooling landscape: the ability to run, track, and coordinate multiple agents across repos without chaos. Its worktree-native approach is elegant, its multi-agent support is comprehensive, and the active community is pushing features like SSH and in-repo worktrees forward fast. If you're running more than one AI agent on a regular basis, Orca is worth a serious look.
Available for macOS, Windows, and Linux. No login required — just bring your own subscriptions.
微信扫一扫,分享到朋友圈
