Orkas

Start with a packaged macOS or Windows installer from the official Orkas site, or build from the public source repository. Linux is supported through source builds. Configure a supported model provider with your own API key or a compatible local model endpoint; model availability and usage costs depend on that configuration.

Give Commander a goal through one chat and supply the relevant files. Commander turns the goal into a plan and coordinates specialist agents in parallel or sequence. For example, a user can request research, a comparison and a presentation draft as connected tasks; this is a proposed workflow, not a measured result. Reusable agents, skills and MCP tools can be configured for the workflow. VideoStudio is built into Orkas for video work.

Authorize only the files, tools and services needed for the task. Review proposed actions and generated results before using them. The local-first workspace can call external model providers and authorized services; it is not a promise that every operation stays offline. Orkas is a desktop AI workforce, rather than a native Grok Bot, ChatGPT GPT or MCP server. The core app is free; model calls and optional services may incur charges.