Orkas Orkas
Home AI local-first Agent
Local-first by design

An AI local-first agent
for work you want to own.

Orkas is an open-source desktop AI client where workspace data stays local by default. Use optional managed official models or connect your own provider; your own-provider calls go directly to it.

Local workspace BYO LLM keys No model proxy
What local-first means

The control plane lives on your machine

For an AI agent, local-first is not only a storage choice. It changes who owns the workspace, the model keys, the generated files, and the reusable workflows you build over time.

Files stay on-device

Use local files, documents, code, notes, and outputs without making Orkas the default cloud host for your workspace.

  • files
  • workspace

Keys remain yours

Connect OpenAI, Claude, Gemini, or other providers with your own account through OAuth or API key.

  • OAuth
  • API key

Choose the model path

Official models use Orkas's managed service. Calls using your own provider go directly from your machine to that provider.

  • managed
  • BYO direct

Workflows compound locally

Agent configs, memory, and reusable sub-agents become local assets you can keep improving.

  • agents
  • memory
Local-first vs cloud-first

A different default for sensitive work

01

Install locally

Run the client on macOS or Windows instead of starting inside a hosted workspace.

02

Choose models

Use optional managed official models or connect your own provider through OAuth or an API key.

03

Direct agents

A lead agent recruits sub-agents and skills. Approval modes prompt for sensitive local operations; routine operations may run without prompting.

04

Own outputs

Generated files, agent runs, and reusable workflows remain yours by default.

Where it fits

Useful when the context is valuable or private

Private documents

Ask PDFs, notes, and files without moving the library into a hosted chatbot.

Codebases

Let repo-aware agents inspect files, reproduce bugs, edit code, and run tests locally.

Recurring workflows

Turn repeat work into sub-agents and keep those patterns in your own workspace.

FAQ

AI local-first agent questions

What is an AI local-first agent?

It is a desktop AI client where core workspace data stays on your machine by default. Model access can use optional managed official models or a provider you connect yourself; calls using your own provider go directly to it.

Does Orkas upload my data or proxy model traffic?

Core workspace data stays on your machine by default. With your own provider, encrypted keys stay local and model calls go directly to that provider; official models use Orkas's managed model service. Optional sync stores only the workspace data you choose to sync.

Is local-first the same as fully offline AI?

No. Local-first means the workspace and control plane are local by default. You can still use cloud models with your own key; only that model request needs the network.

Who should use an AI local-first agent?

People working with sensitive files, codebases, private notes, credentials, or repeatable workflows who want ownership of data, model keys, and long-term outputs.

Try an AI local-first agent on your desktop

Open source, local-first, and built for people who want to own their workspace and model keys.