Anthropic has acquired AI startup Vercept to strengthen its capabilities in “computer use,” a rapidly advancing area that enables AI systems to operate software applications much like a human user.

The acquisition aims to expand the capabilities of Anthropic’s Claude models, which are increasingly being used for complex, multi-step tasks such as writing and executing code across entire repositories, synthesising research from numerous sources and managing workflows across tools and teams.

Computer use allows Claude to interact directly with live applications, navigating interfaces, filling out forms, managing spreadsheets and switching between browser tabs rather than relying solely on static code-based instructions. This enables the AI to complete tasks that would otherwise be difficult or impossible through code alone.

Vercept was founded on the premise that making AI genuinely useful for complex task execution requires solving difficult perception and interaction challenges. The startup’s co-founders — Kiana Ehsani, Luca Weihs and Ross Girshick have focused on building systems that can both “see” and act within the same software environments humans use daily.

Anthropic said Vercept will wind down its external product in the coming weeks, with the team joining the company to further advance Claude’s computer use capabilities.

The acquisition follows the recent launch of Claude Sonnet 4.6, which demonstrated significant improvements in real-world computer interaction tasks. According to Anthropic, its Sonnet models improved their performance on OSWorld, a widely used benchmark for AI computer use, from under 15 per cent in late 2024, when the feature was first introduced, to 72.5 per cent currently. Sonnet 4.6 is now approaching human-level performance on tasks such as navigating complex spreadsheets and completing web-based workflows.

Vercept marks the latest addition to Anthropic’s expanding roster of acquired teams, following its acquisition of Bun. The company said it targets teams whose technical ambitions align with its own and whose work reflects its focus on safety, rigor and responsible AI development.

With the addition of Vercept’s expertise, Anthropic is positioning Claude to take on increasingly sophisticated, real-world digital tasks, moving closer to AI systems capable of operating seamlessly across the software environments that underpin modern work.