- Hugging Face has started an AI tool to navigate the web on your behalf
- Open computer agent uses a real web browser to complete directions or complete tasks like ticket booking
- Agent and its open-source demo can see what is on the screen, click the button, fill the form, and transfer step by step through works like human
Hugging Face has started its own on the increasing number of semi-independent AI agents that can work online for people. New and free (if limited) open computer agent is like being a personal accessory living inside your web browser.
A part of the company’s ongoing “Smolgants” initiatives, the open computer agent can engage with websites and apps like you, which handles an invisible mouse and keyboard. AI can open a browser, type things in forms, click the button, and more. Ask to find directions, and it will go to Google Maps, enter the original and destination, and will show you the way like a duty -digital square.
You can try it yourself with a live demo. Fair warning, its popularity is causing some delays and errors due to a backlog.
We are launching computer using computers in smolagents! 🥳-> As vision models become more capable, they are able to power complex agentic workflows. Especially Qwen-Vl models, which support the underlying grounding, ie its coordinates the ability to detect any element in an image, thus… pic.twitter.com/mi8muwzkisMay 6, 2025
Agent AI
Open computer agent is a different philosophy of an idea that has given birth to similar devices such as operator of openiAI, browser usage, proxy 1.0 and opera operator of opera. Like those devices, hugging the AI agent of the face is about being an active partner rather than a passive source of all information.
Like the use of a browser, the open computer agent is an open-source, which means that one can see how it works and constructs on top of it, or at least it makes it tweezed for cases of niche. The agent is the beginning of some more flexible, not a finished product with a million legal disconnection. This also means that the demo is actually, not a display, not a polish package. This can make things wrong and you need to jump for login and captcha tests.
Booking tickets, checking store hours, searching, looking at the direction -guidelines, and clicking through the menu are all things that many people want to be able to do with single natural language prompts. It is one thing to ask how to find cheap flights, it is one thing to ask. It is to visit a travel website to see a device, scroll via listing, and try to click on “Book Now”.
It can be defective and can be far from attractive, but the open computer agent represents an approach to AI that can now be common as a universal AI image generator.
