Andrew Ng has unveiled OpenWorker, an open-source desktop agent designed to complete tasks such as document creation, calendar updates, and messaging, offering a private and customisable alternative to cloud-based AI assistants.
Andrew Ng has released OpenWorker, an open-source desktop agent designed to do more than chat. According to the project site and GitHub repository, it is built to carry out everyday tasks on a user’s machine and return finished output, such as documents, calendar changes and Slack replies with real data.
The project is positioned as a local-first tool. OpenWorker runs on the user’s desktop, which the developers say helps keep data private while still allowing the agent to work across files, terminals and connected applications. It also supports a bring-your-own-model approach, so users can supply their own API keys or connect local models rather than being locked to one provider.
That design marks a clear shift from the familiar chatbot format. Rather than generating advice or drafts alone, OpenWorker is aimed at producing deliverables that can be used directly in workstreams. The project materials say it can handle polished documents, inbox and calendar tasks, and messaging across tools, with support for a range of integrations. The code is released under the MIT licence, allowing modification and redistribution.
OpenWorker is also intended for developers who want to inspect and extend the system. The GitHub repository provides the full source code, documentation, installation instructions and ongoing issue tracking, indicating an active development process with contributions and updates still in progress. That open setup gives the project a different profile from proprietary assistant products, particularly for teams that value transparency and customisation.
The broader significance is that OpenWorker reflects growing demand for agents that complete work rather than merely discuss it. In practice, that means the product is being framed less as a conversational interface and more as a desktop worker that can be directed, verified and improved over time. For users comfortable managing their own models and tooling, it may offer a practical alternative to cloud-bound assistants.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
