Orca
Run Codex, Claude Code, OpenCode or Pi side-by-side — each in its own worktree, tracked in one place.
About
Orca is a desktop app for orchestrating fleets of coding agents. Fan one prompt across several agents at once, each in its own isolated git worktree, then compare results and merge the winner — all from a single dashboard. A mobile companion app lets you monitor agents and send follow-ups from your phone, with notifications when an agent finishes.
Under the hood it ships Ghostty-class WebGL terminals with infinite splits and a design mode that pipes a clicked UI element's HTML, CSS, and cropped screenshot straight into an agent prompt. The key detail: it runs on your own subscriptions to Codex, Claude Code, OpenCode, or Pi rather than bundling model access.
What it can do
- Parallel agent runs in isolated git worktrees with result comparison
- Mobile companion app with finish notifications and remote follow-ups
- Ghostty-class terminals with WebGL rendering, infinite splits, persistent scrollback
- Design mode: click UI elements to inject HTML/CSS/screenshots into prompts
- Works with Codex, Claude Code, OpenCode, and Pi under your own subscriptions
Setup
- Download the app for macOS, Windows, or Linux from onorca.dev/download
- Install the mobile companion (iOS App Store or Android APK) to steer agents from your phone
- Connect your existing agent subscriptions (Codex, Claude Code, OpenCode, Pi)
- Open a repo, fan a prompt across agents in parallel worktrees, merge the winner
Prerequisites
Desktop OS (macOS, Windows, or Linux); your own paid subscriptions to the coding agents you want to orchestrate — Orca provides no model access itself.
Use cases
Works with
Trust score — 5/10
Scored 8/10 (METRICS tier, metrics-v1). Evidence: ~94.5k stars, MIT license, very active (pushed Oct 17 2026), ships on all three desktop platforms plus iOS and Android, with docs and a Discord community. Docked for: (1) it is an orchestrator only — you still pay for every agent subscription it drives, so the total cost stacks; (2) as a young app handling repo write access across parallel worktrees, worktree hygiene and merge conflicts are the user's problem. License: MIT.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 9.5/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · 94,511 GitHub stars; 6,042 forks; trending across language pages
- · pushed 2026-10-17; multi-platform releases (macOS/Windows/Linux, iOS, Android); docs + Discord
- · manual review: orchestrator only — no model access included, subscriptions stack; parallel-worktree merges are the user's problem
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.