Orca is a desktop and mobile development environment that orchestrates multiple CLI coding agents in isolated parallel git worktrees. Developers must bring their own subscriptions for underlying coding agents, and inference depends on those external agent services.
Project overview
Orca addresses the difficulty of running and comparing multiple CLI coding agents in parallel by isolating each run in its own git worktree and tracking them from a single application that includes a mobile companion.
Project type
AI Agent · AI Coding
Use cases
Coding & Development
Deployment
Refer to project documentation
License
MIT
Best for
Developers and AI engineers who use CLI coding agents and need to run multiple agents in parallel isolated environments.
Teams or individuals tracking pull requests and issues from GitHub and Linear directly within their coding agent environment.
Key capabilities
Run multiple agents side-by-side, each in its own isolated git worktree, allowing comparison and merging of results.
Monitor and steer agents from a phone, get notifications, and send follow-ups.
Provides Ghostty-class terminals with WebGL rendering, infinite splits, and persistent scrollback.
Click UI elements in a Chromium window to send HTML, CSS, and cropped screenshots into an agent's prompt.
Browse pull requests, issues, and project boards directly within the application.
Run agents on remote machines with full file editing, git, and terminals, including auto-reconnect and port forwarding.
Drop comments on diff lines and send them back to the agent for review and editing.
Allows agents or users to script workflows using commands like `orca worktree create`, `snapshot`, `click`, and `fill`.
Limitations and risks
Orca only supports CLI coding agents; non-CLI tools are not supported.
Users must bring their own paid subscriptions for the underlying coding agents.
The data boundary requires external services. Running Orca locally orchestrates the CLI agents, but inference depends on external coding agent requirements.
Getting started
Installers are available for direct download via macOS, Windows, and Linux. The difficulty is rated as easy due to the availability of direct installers and package manager support.