Orkas puts an AI agent team on the desktop, with projects that carry goals, rules and memory across chats
Synopsis
Orkas-AI has released Orkas, an open-source (MIT), local-first multi-agent desktop app: a user tells the Commander a goal, it finishes many tasks on its own, and for larger work it plans steps and brings in specialist agents plus existing coding and personal agents such as Claude Code and Codex. A project holds the goal, rules, memory and files so each new conversation resumes where the last stopped; to-dos split work between people and agents, Auto-advance lets the team take the next item, scheduled automations repeat routine work hourly to monthly, data stays on the computer, and finished work returns for review.
Interpretation
The desktop AI assistant shifts from answering one prompt to running an agent team around an ongoing project. Previously an assistant typically responded to a single prompt, whereas Orkas stores the goal, rules, memory and files in a project so every new conversation picks up where the last one left off. Orkas-AI released it as an open-source (MIT), local-first multi-agent desktop app, shipping 16 specialists with the app and offering 30+ agents in an agent marketplace.
The Commander dispatches the work: it finishes many tasks itself and, for bigger jobs, splits them into steps handed to the right specialist before combining results. The division of labor extends beyond built-in roles, since the Commander can bring in the user's existing Claude Code, Codex, OpenCode, OpenClaw and Hermes-Agent to work on the same job. The stated example turns market research into a report, a slide deck and a launch video; a user can also describe a needed role and have the Commander build a custom agent.
A project to-do list splits work between people and agents, and with Auto-advance on the team takes the next item by itself and brings finished work back for review. Scheduled automations repeat any task hourly, daily, weekly or monthly, and from v1.9 each automation catches up its most recent missed run within 12 hours when the computer wakes or Orkas starts. Orkas runs on the user's computer with data kept local; connectors cover Gmail, Google Calendar, Sheets, Drive, Notion, Slack, Shopify, Stripe and more, and by default Orkas asks before it sends.
Perspective
The demonstrated scope is a macOS and Windows desktop app with 500 credits for the built-in Orkas model, or keys for Claude, OpenAI, Gemini, DeepSeek, Kimi, GLM, Qwen, MiniMax and Doubao or a local OpenAI-compatible endpoint; reference-video recreation works best for motion-graphic explainers, SaaS or app demos and product ads built from existing footage.
Reference-video recreation does not guarantee exact camera moves, human performances or frame-by-frame matching, social-platform share-link downloading is not built in, and text baked into AI-generated footage requires changing that footage; these are conditions listed in the product description.