A team at Hugging Face has released a freely available, cloud-hosted computer-using AI “agent.” But be forewarned: It’s quite sluggish and occasionally makes mistakes.
Hugging Face’s agent, called Open Computer Agent, is accessible via the web and can use a Linux virtual machine preloaded with several applications, including Firefox. Similar to OpenAI’s Operator, you can prompt Open Computer Agent to complete a task — say, “Use Google Maps to find the Hugging Face HQ in Paris” — and sit back as the agent opens the necessary programs and figures out the required steps.
Open Computer Agent can handle simple requests well enough. But more complicated ones, like searching for flights, tripped it up in TechCrunch’s testing. Open Computer Agent also often runs into CAPTCHA tests that it’s unable to solve.
You’ll also have to wait in a virtual queue to use Open Computer Agent — a queue seconds to minutes long, depending on demand.
Of course, the Hugging Face team’s goal wasn’t to build a state-of-the-art computer-using agent. Rather, they wanted to demonstrate that open AI models are becoming more capable — and cheaper to run on cloud infrastructure.
“As vision models become more capable, they become able to power complex agentic workflows,” Aymeric Roucher, a member of the agents team at Hugging Face, wrote in a post on X. “[Some of these models] support built-in grounding, i.e. [the] ability to locate any element in an image by its coordinates, [and] thus [can] click any item [in a virtual machine].”
While it’s far from perfect, agentic technology is attracting increasing investment as enterprises look to adopt it to boost productivity. According to a recent KPMG survey, 65% of companies are experimenting with AI agents. Markets and Markets projects that the AI agent segment will grow from $7.84 billion in 2025 to $52.62 billion by 2030.
Techcrunch event
Berkeley, CA
|
June 5
