Orkas

Orkas is an open-source, local-first desktop AI workforce where a Commander LLM coordinates specialist agents in parallel or in series through one chat, keeping files and keys on your machine while supporting managed models or bring-your-own keys.

Orkas is an open-source, local-first multi-agent desktop client where a super-powered Commander coordinates specialist agents to complete complex work together. You describe a goal in a single chat, the Commander breaks it into steps, handles general parts itself, and routes specialist stages to agents like DeepResearcher, ContentWriter, PptMaker, ImageStudio, VideoStudio, OfficeWorker, ProductDeveloper, UIDesigner, and SeoGeoAgent. The result lands as finished files on your disk that you can use immediately or keep editing. The product is local-first by default: chats, project history, knowledge material, and generated files stay on your device, and encrypted credentials protect API keys locally. You can use managed official models that route through the Orkas-managed service, or connect your own providers via OAuth or API keys, including Anthropic, OpenAI, Gemini, DeepSeek, Kimi, GLM, MiniMax, Doubao, and OpenRouter. Orkas can also drive installed coding CLIs as local sessions, among them Claude Code, Codex, OpenClaw, OpenCode, and Hermes. Agents can self-evolve through reflection and skill crystallization, and reusable agents and skills are available in a marketplace of 100-plus ready capabilities. Released under the MIT license at github.com/Orkas-AI/Orkas with 2k-plus GitHub stars and weekly releases, Orkas targets indie builders, one-person businesses, and small teams that want complex work to become a reusable team workflow rather than a one-off prompt. The desktop app is free with 500 bonus credits, and paid plans add monthly credits, watermark-free output, cloud storage, and multi-device sync.