Orkas
Orkas coordinates specialist AI agents from a desktop chat, with local project storage and a choice of managed or user-configured model providers.
Project links
Project Note
A closer look at Orkas
A desktop app that plans multi-step AI tasks and coordinates specialist agents for research, writing, files and software work.
Research, writing and slides in one task
Orkas is built around work that has several parts. Its README uses a competitor-research request that ends in a slide deck: a research agent gathers sources, a writing agent prepares the report and a presentation agent builds the slides. A Commander coordinates those steps from one desktop conversation. This is an example supplied by the project, not a workflow ProjectRadar ran.
The useful idea is keeping the intermediate work beside the final files. Someone preparing a briefing could review the source material before accepting the report and deck, instead of moving each result between separate chats. Orkas also documents specialist agents for office files, software work, images and video, and support for external local agents. Output quality still depends on the models, tools and review given to the task.
Local storage and remote requests
The public repository describes a MIT-licensed desktop app, with Mac and Windows installers and a Linux option that runs from source. The packaged app asks users to sign in; the repository describes a source build without that account layer. ProjectRadar inspected the README and official pages without installing either version.
Orkas says project data is stored locally by default. Its privacy policy makes the exceptions clear: direct provider requests send relevant content to the chosen provider, while managed AI and tool features route content through Orkas and upstream services. Optional sync uploads selected product data. A local workspace therefore does not make a cloud-model request private to the computer.
Agents can use files and tools with the access granted to the app. Before bringing in client documents or connecting another service, the practical question is which agent can read or change what, and where its model requests go. The public security description is not an independent audit.
Free app, separate usage costs
The pricing page offers core access without a subscription and 500 welcome credits on desktop sign-in. Managed model and media use spends credits; a user-configured provider can have its own charges. Paid plans add monthly credits, cloud storage and sync. Some advanced capabilities are marked planned, so the current plan details matter more than the breadth of the homepage examples.
For reference
Project facts
The published details and source context, together in one place.
- Product
- Orkas
- Category
- AI & Skills
- Stage
- Launched
- Public Context Check
- Public pages were checked Oct 2, 2026.
Project note details
Dates and editorial links
Published
October 4, 2026
Updated
October 4, 2026
Page type
Project Note
Status
Published note
Publication history
Note history
Publication and substantive update dates for this public note.
- Oct 4, 2026
- Project note published.
- Oct 2, 2026
- Public-page information added to the note.
More projects
Related Project Notes

Vecbase
Vecbase lets teams assign work to AI agents, share files and skills, schedule tasks, and watch tool activity in one workspace.

Tale
A shared project workspace for task planning, agent delegation and review of reports and files.

Brain Scanner
A hosted workspace presenting project graphs, recorded agent work, files, reports, and task queues.
Building something useful?
Submit your project for free. If selected, it gets a profile that explains what it does and links to your work. Publication isn't guaranteed.
Submit a project