Orkas

Orkas is an open-source, local-first multi-agent AI desktop client for macOS and Windows. A super-powered Commander coordinates specialist agents to complete complex work together.

Use Orkas as a coordinated AI team for complex work: research markets, turn findings into reports or slides, read PDFs with citations, fix bugs and open PRs, analyze data, create apps or videos, organize inbox and calendar work, connect external tools, and more, then turn repeated work into reusable agents and skills.

ChatGPT is a hosted assistant, while Codex, Claude Code, and Cline focus on coding sessions. Orkas is an open-source, local-first AI workforce where a Commander coordinates specialist agents to complete complex work together.

No. Orkas can provide managed official models that run through Orkas-managed model service. You can also connect your own providers through OAuth or API keys, including OpenAI, Claude, Gemini, DeepSeek, Kimi, and GLM; bring-your-own-provider calls go directly to that provider.

By default, your configs, history, knowledge bases, and generated files stay on your machine. If you bring your own keys, they are stored locally and requests go directly to your provider. Orkas-managed models and optional cloud features use Orkas services only for those selected features.

The desktop app is free and includes 500 bonus credits for managed official models; bring-your-own-provider usage is billed by your provider. Paid plans add monthly credits, multi-device sync, and cloud storage. Remote control and team collaboration are planned.

Credits meter the Orkas-managed model service only. If you connect your own provider key, those calls go straight to that provider and are billed by it, so they use no credits. Paid plans include a monthly allowance — 5,000 on Lite, 10,000 on Pro, and 50,000 on Max — and you can also buy lifetime credit packs as a one-time purchase, which work exactly like monthly credits and are available whether or not you subscribe (monthly credits are spent first). Consumption follows each provider's official pricing: roughly 100–300 credits per million input tokens, 2–6 per million cached input tokens, and 200–600 per million output tokens, so repeated context costs far less than it looks.

Yes. The Orkas desktop client is open source under the MIT license at github.com/Orkas-AI/Orkas. You can read the source, build it yourself, and audit exactly what the app sends over the network.

Yes. The open-source desktop edition is free to download and self-host, and it runs without an Orkas account: you add your own provider keys and your conversations, files, and keys stay on your machine. The official release adds the optional hosted services — managed models, multi-device sync, and cloud storage — and those need an Orkas account. One thing worth being clear about: Orkas is a desktop client, not a server, so self-hosting means running the app on your own machine rather than deploying a server for a team.

The official download page provides macOS and Windows builds. The open-source repository also supports Linux, which you build yourself. The interface ships in English, Chinese, Japanese, and Portuguese, and you can switch language inside the app at any time.