Orca is an open-source Agentic Development Environment (ADE) from Stably that lets a single developer run a fleet of AI coding agents in parallel, each isolated in its own git worktree and tracked from one place. Released under the MIT license and available as a desktop app for macOS, Windows, and Linux with a mobile companion, Orca reframes the terminal as an orchestration surface: instead of babysitting one agent in one chat, you fan work out across many and merge the winners. ## One Home for Every Coding Agent Orca's core idea is that you shouldn't be locked into a single tool. It runs Codex, Claude Code, OpenCode, and Pi side by side, each with your own subscription, so you can pick the right agent for a task or race several against the same prompt. Every agent runs in an isolated git worktree, which means concurrent agents never stomp on each other's files, and you can review, diff, and merge each result independently. This bring-your-own-agent model keeps you vendor-neutral while consolidating the workflow into one dashboard. ## Parallel Worktrees and Racing The standout capability is fanning one prompt across multiple agents. You can dispatch the same task to five agents, let each work in its own isolated worktree, then compare the outputs and merge the best one. For open-ended engineering problems where the first attempt is rarely the best, this parallel-and-select pattern turns agent variance into an advantage instead of a frustration, and the unified view makes it practical to actually evaluate the candidates. ## Terminal, Desktop, and Mobile Orca ships Ghostty-class terminals with WebGL rendering, infinite splits, and scrollback that survives restarts, so the environment feels like a first-class terminal rather than a chat box bolted onto an IDE. A mobile companion app for iOS and Android lets you monitor agents from your phone, receive a notification when an agent finishes, and send follow-up instructions from anywhere — useful for long-running tasks you kick off and check on later. ## Trade-offs and Limitations Orca is an orchestrator, not an agent itself: it does not replace Codex, Claude Code, or the others, and you still need active subscriptions to the underlying agents to get value from it. The parallel-worktree model assumes a git-based workflow, and running several agents at once multiplies token spend and local compute. As a fast-moving, young project it is evolving quickly, so expect rough edges and frequent releases rather than a settled, stable surface. ## Who Should Use This Orca is aimed at developers who already lean heavily on AI coding agents and want to scale past one-agent-at-a-time. If you routinely run multiple agents, want to compare approaches before committing, or need to keep an eye on long tasks from your phone, Orca provides the isolation, tracking, and cross-tool orchestration to make that manageable — all under a permissive MIT license.