Orca
Agent development environment published by Stably AI under the MIT licence, on macOS, Windows and Linux. It runs several command-line agents side by side (Claude Code, Codex, OpenCode, GitHub Copilot, Pi and many more), each in its own Git worktree, with split terminals. You comment on a diff line and the comment goes back to the agent, you open a worktree from a GitHub issue or a Linear task, and an `orca` CLI lets you script the lot. Worktrees can also run over SSH on a remote machine.
Strengths
- One Git worktree per agent, so two agents do not step on each other
- Agent agnostic: Claude Code, Codex, OpenCode, Copilot and any command-line agent
- Comments dropped on a diff line and sent back to the agent, without leaving the tool
- MIT licence, GitHub issues and Linear tasks wired in natively
Limitations
- Oversized as long as you only drive one agent at a time
- A project that ships very often: what you set up today can move quickly
- More parallel agents means more diffs to review and more tokens burned
Best for
- Running the same prompt across several agents and keeping the best result
- A developer already running two or three coding sessions in parallel