OpenCoworker 0.1.0 — first public beta
OpenCoworker is an AI coworker that lives on your desktop. Give it a folder and a task, and it works the way a colleague would — researching, writing, and building real files on your machine, not just chat replies.
What's in this release
- Works in your folders — grant access to any folder (read-only or read-write); everything the agent produces is a real file you keep.
- Bring your own model — connect OpenAI, Claude (Anthropic), or Gemini with your API key, or run fully local with Ollama. Switch models per conversation.
- Built-in viewer — preview the documents, images, and reports the agent creates without leaving the app.
- Automations — schedule recurring tasks ("every morning, prepare a market brief") that run even when you're not around.
- MCP support — extend the agent with Model Context Protocol servers (local or remote), with per-tool approval.
- You stay in control — risky actions ask for your approval first, and your API keys never leave your machine.
- Light, dark, and auto appearance.
Download
| Platform | File |
|---|---|
| macOS (Apple Silicon, M1 or later) | OpenCoworker-macos-arm64.dmg
|
| Windows 10/11 (x64) | OpenCoworker-windows-setup.exe (or the .msi)
|
The macOS app is signed and notarized — it opens like any other app. On Windows, SmartScreen may show a warning on first run (the build isn't Authenticode-signed yet): choose More info → Run anyway.
You'll need an API key from OpenAI, Anthropic, or Google — or a local Ollama install — to connect a model during setup.