Overview
Orca is an open-source Agent Development Environment (ADE) designed for working with a fleet of parallel AI coding agents. It allows developers to fan one prompt across multiple agents in isolated git worktrees, compare results, and merge the winner. Agent-agnostic and cross-platform, Orca supports Claude Code, Codex, Cursor, Copilot, Grok, Devin, and more.
Features
- ✓Parallel worktrees 鈥?fan one prompt across multiple agents in isolated git worktrees
- ✓Mobile companion apps (iOS, Android) for monitoring and steering agents remotely
- ✓Terminal splits with WebGL rendering and persistent scrollback
- ✓Design mode 鈥?click UI elements to send HTML, CSS, and screenshots into agent prompts
- ✓Native GitHub and Linear integration for PRs, issues, and boards
- ✓SSH worktrees with auto-reconnect and port forwarding
- ✓Annotate AI diffs 鈥?comment on diff lines and send feedback back to the agent
- ✓Orca CLI for scripting workflows
Installation
brew install --cask stablyai/orca/orcaPros
- +Agent-agnostic: works with Claude Code, Codex, Cursor, Copilot, and more
- +Parallel execution dramatically accelerates development
- +Rich desktop app with mobile companions
- +MIT licensed 鈥?free and open source
- +Cross-platform: macOS, Windows, Linux
Cons
- −Desktop app 鈥?not a library or framework
- −Requires multiple agent subscriptions to fully utilize
- −Newer project with rapidly evolving features
- −Learning curve for parallel agent workflows
Alternatives
Documentation
Orca
Overview
Orca is an open-source Agent Development Environment (ADE) designed for working with a fleet of parallel AI coding agents. Unlike traditional single-agent tools, Orca allows developers to fan a single prompt across multiple agents running in isolated git worktrees, compare the results side-by-side, and merge the winning implementation.
Agent-agnostic by design, Orca works with Claude Code, Codex, Cursor, GitHub Copilot, Grok, Devin, OpenCode, and any other CLI-based coding agent. It provides a rich desktop application with terminal splits, mobile companion apps, and deep integration with GitHub and Linear for seamless development workflows.
Features
- Parallel Worktrees: Fan one prompt across multiple agents, each in an isolated git worktree. Compare results and merge the winner.
- Mobile Companion: iOS and Android apps to monitor and steer agents, get notifications, and send follow-ups from anywhere.
- Terminal Splits: Ghostty-class terminals with WebGL rendering, infinite splits, and scrollback that survives restarts.
- Design Mode: Click any UI element in a real Chromium window to send its HTML, CSS, and a cropped screenshot into an agent's prompt.
- Native GitHub & Linear: Browse PRs, issues, and boards in-app; open worktrees from tasks without context switching.
- SSH Worktrees: Run agents on remote machines with auto-reconnect and port forwarding.
- Annotate AI Diffs: Comment on diff lines and send feedback back to the agent for iterative improvement.
- Orca CLI: Script workflows with commands like
worktree create,snapshot,click, andfill.
Installation
macOS (Homebrew)
brew install --cask stablyai/orca/orca
Arch Linux (AUR)
yay -S stably-orca-bin
Direct Downloads
Available for macOS (Apple Silicon & Intel), Windows (.exe), and Linux (AppImage) from the releases page.
Quick Start
- Install Orca via the platform-appropriate method above.
- Launch Orca and connect your agent subscriptions (Claude Code, Codex, etc.).
- Create a new project and link it to a GitHub repository.
- Write a prompt describing the feature or fix you want to implement.
- Fan the prompt across 3-5 agents running in parallel worktrees.
- Review the generated diffs side-by-side and merge the best result.
Core Concepts
Parallel Worktrees
Each agent gets its own isolated git worktree branched from the same base. This means agents can work independently without merge conflicts during execution. You review the results and choose which to keep.
Agent Agnosticism
Orca doesn't replace your coding agents — it orchestrates them. You bring your own subscriptions to Claude Code, Codex, Cursor, Copilot, or any other CLI agent, and Orca manages the parallel execution.
Design Mode
When working on UI changes, Design Mode lets you click any element in a real Chromium browser window. Orca captures the HTML structure, CSS styles, and a screenshot, and injects them into the agent's context for precise UI modifications.
Advanced Features
- Remote Execution: Run agents on powerful remote machines via SSH with automatic reconnection.
- Workflow Scripting: Automate repetitive multi-agent workflows using the Orca CLI.
- Diff Annotation: A review workflow where you can annotate specific lines and feed feedback back to the agent.
- Session Recovery: Terminal scrollback and agent state persist across application restarts.
Examples
- Feature Development: Fan a "add user authentication" prompt across Claude Code, Codex, and Cursor, compare all three implementations, and merge the best.
- Bug Fixing: Send the same bug report to multiple agents with different frameworks (React, Vue, Svelte) and evaluate which solution is cleanest.
- Code Review: Use one agent to write code and another to review it, running both in parallel worktrees.
- Refactoring: Test different refactoring approaches simultaneously and compare test results.
Pros
- ✅ Agent-agnostic: Works with all major coding agents — no lock-in
- ✅ Parallel execution: Dramatically accelerates development by comparing multiple solutions
- ✅ Rich desktop experience: Native app with WebGL terminals, mobile companions, and GitHub integration
- ✅ MIT licensed: Free and open source
- ✅ Cross-platform: macOS, Windows, and Linux support
Cons
- ❌ Desktop application: Not a library or framework — requires the Orca app
- ❌ Multiple subscriptions needed: Full potential requires multiple agent subscriptions
- ❌ Rapidly evolving: Features change frequently as the project matures
- ❌ Learning curve: Parallel agent workflows require new development patterns
When to Use
Use Orca when:
- You want to compare multiple agent implementations of the same task
- You're building a complex feature and want to explore different architectural approaches
- You manage a team of agents and want a unified interface for orchestration
- You need mobile access to monitor and steer agent tasks remotely
- You want to script and automate multi-agent workflows
