An open-source AI assistant
you can actually inspect and own.
Orkas is a local-first desktop AI assistant and multi-agent workspace. Choose optional managed official models or bring your own provider, inspect the source on GitHub, and direct a team of agents instead of relying on one hosted chatbot thread.
A visible client, direct model access, and local control
People searching for an open-source AI assistant usually want more than a chat box. They want auditability, model choice, and a workspace that remains useful even when a hosted service changes.
Inspect the client
Read the source, understand the data model, and see how agents, skills, and local files are orchestrated.
- source
- audit
Choose your model path
Use optional Orkas-managed official models, or connect OpenAI, Claude, Gemini, or another provider with your own account. With your own provider, traffic goes directly to it and is not proxied through Orkas.
- managed models
- OAuth
- API key
Keep the workspace local-first
Chats, agent configs, generated files, and keys stay on your machine by default.
- local-first
- BYO keys
Build with agents and skills
Create reusable sub-agents and skills, then summon them from one conversation.
- agents
- skills
A desktop workspace instead of a hosted assistant alone
Run it locally
Use a native desktop app on macOS or Windows.
Connect models directly
Your model calls go to the provider you choose.
Direct a team
A lead agent recruits sub-agents and skills for larger work.
Own the workflow
Repeatable work becomes your reusable agents and local assets.
Open-source AI assistants, compared
These tools solve different problems. The last row is where the others are the better answer.
| Orkas This site | Jan | AnythingLLM | LM Studio | PyGPT | |
|---|---|---|---|---|---|
| License | ●Open source (MIT) | ●Open source | ●Open source | Proprietary, free to use | ●Open source (MIT) |
| What it is built for | ●A multi-agent workspace for real project work | Running local models in a clean chat | Private document understanding and RAG | Running and managing local models | A desktop assistant with node-based agent workflows |
| Model access | ●Managed models or your own provider, called directly | Local models, or a provider you connect | Local models, or a provider you connect | Local models | Local models via Ollama, or a provider you connect |
| Agents working as a team | ●Lead agent dispatches sub-agents in parallel or in series | Single assistant | Single assistant with agent skills | Chat only | Workflows you design node by node |
| Recurring work becomes | ●Reusable sub-agents and skills stored locally | Saved threads and prompts | Document workspaces | Saved chats and model presets | Agent workflows built in its node editor |
| Simplest way to just run a local model | Possible, but not the point of the app | ●Yes, this is its core job | Yes, alongside documents | ●Yes, this is its core job | ●Yes, via Ollama |
If your goal is to download a local model and chat with it, Jan and LM Studio do that more directly, and AnythingLLM is purpose-built for question-answering over a document set. PyGPT is the closest single-app alternative. Both run on Linux (PyGPT natively; Orkas through its open-source edition, which you build from source, while the official builds cover macOS and Windows), and PyGPT wins on one thing Orkas does not do: it drives local models through Ollama. Orkas is the better fit when the assistant has to carry out multi-step work across your real files with a team of agents rather than answer inside one thread.
Choose Orkas when openness and ownership matter
Auditability
You can inspect how the desktop client works and reason about the trust boundary.
Model choice
Switch between providers and accounts without changing your workspace.
Long-term control
Your agents, skills, configs, and outputs are not locked inside a hosted chat product.
Open-source AI assistant questions
Is Orkas an open-source AI assistant?
Yes. Orkas is an open-source desktop AI assistant and multi-agent workspace. The source is available on GitHub, and users can use optional Orkas-managed official models or connect their own providers through OAuth or API keys.
Is Orkas an open-source ChatGPT alternative?
Orkas can replace ChatGPT for users who want a local-first desktop workspace, open source code, optional managed models or their own provider, and multi-agent workflows. ChatGPT remains stronger if you want a fully hosted assistant with no setup.
Does Orkas provide the AI model?
Orkas offers optional managed official models. You can also connect OpenAI, Anthropic Claude, Google Gemini, or another provider through OAuth or an API key. With your own provider, traffic goes directly to it and is not proxied through Orkas.
Who should choose an open-source AI assistant?
Choose an open-source assistant when auditability, local control, model choice, and long-term ownership matter more than a zero-setup hosted chatbot.
Which open-source AI assistant should I choose?
Pick by the job, not by the label. Jan and LM Studio if you mainly want to download a model and chat with it. AnythingLLM for question-answering over a document set. PyGPT for one desktop app with node-based agent workflows. OpenClaw if you want a self-hosted assistant reachable through messaging channels. Orkas when the work needs a team of agents carrying out multi-step tasks across your real files, rather than one thread answering questions.
Is an open-source AI assistant the same as running the model locally?
No, and mixing the two up is the most common mistake in this category. Open source describes the client: you can read it, fork it, and see what it does with your data. Where the model runs is a separate decision. Orkas is open source and local-first — chats, agent configs and keys stay on your machine — but it does not ship a local LLM. It calls optional managed official models, or the provider you connect, and with your own provider the call goes straight there. If you want the model itself on your machine, Jan, LM Studio, or PyGPT through Ollama do that directly.
Try an open-source AI assistant on your desktop
Free and local-first, with optional managed models or your own keys. Install the desktop app and direct your AI team by chat.