Andrew Ng open-sources OpenWorker, a local-first desktop agent that can deliver finished work
量子位 · wechat · 2026-07-25
Andrew Ng open-sources OpenWorker, a desktop AI agent that ships finished work
Andrew Ng has launched OpenWorker, an open-source, MIT-licensed desktop agent designed to do more than chat: it can take a task like “prepare a customer brief” and actually produce the finished output across files, calendars, Slack, and other workplace tools.
What it does:
- Connects to 25+ tools, including GitHub, Slack, Jira, Notion, Linear, HubSpot, Outlook, Gmail, and Google Calendar.
- Can also access local files and the terminal.
- Supports more tools through MCP.
- Can run scheduled workflows such as daily briefings, weekly reports, and channel monitoring.
OpenWorker is built around four principles: open source, local-first, privacy protection, and model agnosticism.
- It works on Mac, with a Windows version coming soon.
- Users can bring their own API key for models such as GPT, Claude, Gemini, or run local/open-weight models through Ollama.
- By default, agent loops, conversation history, connector tokens, and model keys stay on the user’s device unless a chosen service or integration sends data out.
A key safety design is that the agent pauses for approval before sending messages, editing calendars, writing to external tools, or running terminal commands. If the user is away, these actions are queued for confirmation instead of being executed blindly.
Ng’s broader message is clear: an AI agent should not stay trapped in a chat window; it should understand context, produce deliverables, and return control to the human at critical steps.
Related event: Andrew Ng Releases Open Source Agent OpenWorker(6 posts)→
More from coding & agent
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- Anthropic researcher: 99% of engineers now run swarms of 300+ self-improving agents — AlishaOutridge · 2026-09-11
- Gergely Orosz: Shipping 10x PRs With AI Agents, Sites Fill With Small Regressions — ducha_aiki · 2026-09-11
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Astra storyboards plus Minimax H3 per-shot generation boost video success rates — Hailuo_AI · 2026-09-11
- Codex tip: use Sol with Astra and Luna sub-agents to save usage — pvncher · 2026-09-11