Orkas

Overview

Orkas is a desktop application for coordinating AI agents through chat. Its Commander plans tasks, handles work directly and delegates to specialists in sequence or in parallel. The app includes agents for research, writing, presentations, development and other document or media tasks.

The README describes local storage for conversations, files and credentials. Cloud model calls still send data to the selected provider. You can also configure a compatible local model endpoint or run installed external CLI agents, including Claude Code and Codex, as local sessions.

How it works

1. Configure a model provider with your own credentials, or add a compatible local endpoint.
2. Describe a goal in chat. The Commander decides which parts to handle and which specialists to involve.
3. Agents use their own skills, memory and tools to complete assigned work. The Commander can consume their results for later steps or hand over the final response to a specialist.
4. Review the returned results and files before using them. External CLI integrations require the corresponding CLI to be installed and configured.

Examples

The README demonstrates researching five competitors, writing a report and turning it into a presentation. DeepResearcher gathers sources, ContentWriter drafts the report and PptMaker builds the slides.

Other documented uses include turning a product specification into development tasks and working with uploaded documents for local data analysis.

Notes

Packaged installers for Windows and macOS require sign-in. The source build has no account layer. Your model provider may charge for usage; the packaged app also offers an optional Orkas model billed in credits.

This directory has not independently verified installation or task execution. On October 13, 2026, the latest native-dependency CI workflow failed during package installation on all five platform jobs, including Windows, while the Linux source smoke workflow passed. The cause of those installation failures was not confirmed. See the CI run before relying on a source installation.

The repository's main license is MIT, with the notice Copyright (c) 2026 Orkas contributors. Bundled components have separate license notices; preserve those notices if you redistribute them. Listing Orkas here does not imply endorsement by its authors.

Installation

You need Git and Node.js 22.12 or newer to run from source. The README recommends Node.js 24 LTS for native-dependency troubleshooting. Supported systems are Windows 10+, macOS and glibc 2.34+ Linux on x64 or arm64. Alpine and other musl-based Linux distributions are unsupported.

1. Clone the repository and enter its directory:

   git clone https://github.com/Orkas-AI/Orkas.git
   cd Orkas
   

2. On Windows, launch it from PowerShell:

   .\run.cmd
   

   On macOS or Linux, use:

   ./run.sh
   

3. Allow the first launch to install dependencies and download runtime resources, including the embedding and speech models. Python 3 and a C/C++ build toolchain may be needed when a native package must compile locally.
4. Configure your provider credentials or local model endpoint in the app. For a custom endpoint, the README specifies Account settings, Models, Chat, Add model, then Custom. Supply the protocol, base URL and model ID.
5. Run a small task and inspect its response and any generated files. Confirm that the selected model responds before trying a longer workflow.

For native-dependency errors, the README documents npm run native:repair. Use npm run native:check to check dependencies without installing them. Windows and macOS users can instead choose a packaged installer from the official download page, subject to its sign-in requirement.