XNews

Andrew Ng's OpenWorker: Open-Source Desktop AI Agent

Andrew Ng's OpenWorker: Open-Source Desktop AI Agent

Andrew Ng announced OpenWorker, an open-source desktop AI agent that goes beyond chat to deliver completed work such as polished documents, calendar updates, and Slack messages. It connects to 25 tools including GitHub, Slack, Jira, Notion, Gmail, and Google Calendar, and asks for confirmation before performing sensitive actions. OpenWorker runs on Mac with Windows support coming, supports multiple models (OpenAI, Claude, Gemini, DeepSeek, Kimi, Qwen, Ollama), and keeps data local except when sent to the user's chosen LLM provider. The project aims to provide an open, privacy-preserving, model-independent alternative for AI coworkers. The source code is available on GitHub.

Key insights

OpenWorker fills a gap for a privacy-focused, model-agnostic desktop agent that doesn't lock users into a single ecosystem

Unlike platform-specific assistants, OpenWorker lets users bring their own API keys and choose from a wide range of models, including local options via Ollama. Data stays on the user's machine unless explicitly sent to a chosen LLM provider or integration, addressing concerns about vendor lock-in and data privacy.

The open-source nature and local-first design address growing concerns about data privacy and vendor lock-in in AI assistants

OpenWorker's code is fully open-source, and conversation records and model keys are stored locally. This transparency and local control contrast with cloud-based assistants that may use data for training or create dependency on a single provider.

By supporting multiple models, OpenWorker allows users to choose the best or most cost-effective LLM for each task, promoting flexibility

The agent works with GPT 5.6 Sol, Claude Fable, Gemini 3.6, and open-weight models like Kimi, GLM, DeepSeek, and Inkling, as well as local Ollama models. This model independence lets users optimize for cost, speed, or capability per task without switching tools.

Source posts