Hugging Face Unveils a Free AI Tool with Operator-Like Agent Capabilities

A team at Hugging Face has launched a free, cloud-based AI “agent” capable of operating a computer — though it’s relatively slow and sometimes prone to errors.

A Web-Based Linux Tool for Automated Task Execution

Known as the Open Computer Agent, the tool is available through a web interface and runs on a Linux virtual machine equipped with various apps, such as Firefox. Much like OpenAI’s Operator, users can instruct it to perform tasks — for example, “Find Hugging Face HQ in Paris using Google Maps” — and the agent will autonomously open applications and carry out the needed steps.

Hugging Face didn’t set out to create the most advanced AI agent; instead, the goal was to show how open-source AI models are growing in capability and becoming more affordable to run in the cloud.

Vision Models Gain Grounding Capabilities for Interactive Virtual Tasks

“As vision models improve, they’re increasingly able to handle complex, agent-like tasks,” said Aymeric Roucher from Hugging Face’s agents team on X. “Some can now perform grounding — identifying the exact location of elements in an image — allowing them to interact with virtual environments by clicking items.”

Although still imperfect, agentic AI is gaining momentum, with many businesses exploring its use to enhance productivity. A recent KPMG survey found that 65% of companies are experimenting with AI agents, and Markets and Markets forecasts the sector will grow from $7.84 billion in 2025 to $52.62 billion by 2030.

Would you like this paraphrased further for a different tone or format?

Read the original article on: TechCrunch

Read more: GPT-4.1 May Be Less Aligned With User Intentions Than Earlier OpenAI Models