OpenWorker

OpenWorker is an open-source desktop AI agent, released by Andrew Ng, built around a simple distinction: instead of returning a chat reply, it hands back a finished deliverable, a polished document, a Slack message with the actual numbers in it, an updated calendar entry, a triaged inbox. Tell it the outcome you want, and it breaks the request into steps, works across local files, the terminal, and connected apps, then checks in for approval before doing anything consequential like sending a message or running a command.

It runs entirely on the user's machine, a Tauri desktop shell over a local Python agent server built on Andrew Ng's aisuite library, so files, conversations, and connector credentials stay local rather than passing through a vendor's cloud. Model access is bring-your-own-key across OpenAI, Anthropic, Google Gemini, DeepSeek, Kimi, Qwen, MiniMax, Mistral, Grok, and other tool-calling models, plus fully local inference through Ollama for people who want zero data leaving the device. Connectors cover 25+ everyday tools, GitHub, Slack, Jira, Notion, Linear, HubSpot, Outlook, monday.com, Gmail, and Google Calendar, plus any MCP-reachable tool, and it can also be triggered directly from a Slack mention or run on a schedule for recurring work like a morning brief or a weekly report.

OpenWorker is MIT-licensed, in open beta, and currently ships for macOS with Windows support in progress, aimed at people who want an agent that finishes real tasks end to end rather than one more chat window to copy answers out of.