Running a single coding agent inside an existing IDE used to be the standard workflow. Stably, the company behind the Orca project, is betting that workflow no longer fits how developers actually build software in 2026. Orca, released as open source under the MIT license, is designed from the ground up to manage multiple coding agents running simultaneously, each in its own isolated git worktree.
The core mechanic is parallel worktrees: a developer can fan one prompt across five agents, let each produce a different implementation, then compare the outputs and merge whichever wins. That pattern treats agents more like competing contractors than sequential assistants. Each agent works independently, without interfering with the others’ working directories, and the developer reviews the results rather than watching one agent proceed step by step.
Orca supports more than 30 named CLI coding agents at launch, including Claude Code, OpenAI’s Codex, Grok, Cursor, GitHub Copilot, Cline, and Devin. The compatibility model is broad by design: if an agent runs in a terminal, Orca will run it. That positioning makes Orca agent-agnostic infrastructure rather than a vehicle for any one provider’s ecosystem.
The feature set goes beyond terminal management. A Design Mode lets developers click any UI element in an embedded Chromium window and send that element’s HTML, CSS, and a cropped screenshot directly into an agent’s prompt, shortening the feedback loop for visual changes. An annotation layer lets developers drop comments on any diff line and return the annotated diff to the agent, keeping the review loop inside the same application. GitHub and Linear are natively integrated, so developers can open a worktree directly from a task or pull request without switching context.
A mobile companion app, available on iOS and Android, lets developers monitor agent status and send follow-up instructions from a phone. That feature targets a specific pain point in long-running agent sessions: the developer who queues several tasks and wants to know when they finish without being chained to a laptop.
The README, published on GitHub by Stably on June 24, does not include adoption data, star counts, or independent benchmark comparisons with other orchestration approaches. The release announcement is self-described and the product claims in it are the company’s own. Whether the “100x builder” framing reflects genuine productivity gains at production scale is not addressed by the material.
The harder strategic question is whether agent fleet management survives as a standalone product category. VS Code, JetBrains, and Cursor all have strong incentives to absorb multi-agent workflows natively, and Anthropic has already built Claude Code with worktree support. A standalone ADE (the term Stably uses) that works across all agents has a clear proposition today, when the agent ecosystem is fragmented across CLIs and subscriptions. That proposition narrows as the major IDEs consolidate multi-agent features into their core surfaces.
Stably’s claim to differentiation rests on neutrality. Orca is not selling one agent; it is selling the orchestration layer above all of them. The MIT license and open-source distribution lower the adoption barrier for teams that cannot commit to a vendor-specific workflow. Whether that neutrality is durable depends on how quickly incumbent IDEs move.
Development teams running three or more agents in parallel now have an explicit architecture decision to make: build orchestration into their existing IDE setup, use a purpose-built tool like Orca, or wait to see which IDE vendors ship native fleet management first.
Published as an open-source project on GitHub by Stably (stablyai/orca repository), June 24, 2026.
