Anthropic acquires Vercept, a startup specializing in computer control
Anthropic has acquired Seattle startup Vercept, known for developing advanced agentic tools. The company's key product was an AI agent for computer control: it
AI-processed from TechCrunch; edited by Hamidun News
Anthropic has acquired Seattle-based startup Vercept, which specialized in developing agent tools for computer management. The deal closed at a challenging moment: just before its completion, Meta managed to poach one of the company's founders. Despite this, the acquisition took place — and it unequivocally signals where Anthropic intends to move in the coming years.
The race for so-called "computer use" — the ability of AI agents to independently work with application interfaces the way a living human does — unfolded among leading technology companies back in 2024. OpenAI introduced an operator capable of navigating a browser and filling out forms. Google integrated similar capabilities into its Gemini ecosystem. Anthropic didn't stay on the sidelines: in autumn 2024, the company announced its own computer use feature for Claude 3.5 Sonnet, which allowed the agent to control the cursor, click elements, and enter text. However, in terms of implementation maturity, Anthropic lagged behind competitors — and it is precisely this problem that the Vercept acquisition is meant to solve.
The Seattle startup made a bet on what most market players merely declare: creating agents capable of working inside real applications under conditions as close as possible to everyday user practice. Vercept's key product is an AI agent that not only navigates web pages but also interacts with desktop programs: enterprise software, utilities, complex interfaces where standard API integrations are simply unavailable. Essentially, it's an agent that "sees" the screen and "acts" with its hands — just like an employee at their laptop. The difference between an attractive demonstration and a reliable industrial solution is enormous here, and Vercept, apparently, has progressed considerably further than many.
Meta's attempt to poach one of Vercept's founders before the deal closed speaks volumes. Major technology corporations have long competed not just for finished products, but for specific people with rare expertise. Agent systems for computer management require a specific combination of knowledge: understanding of computer vision, processing of user interfaces, planning of multi-step tasks, and reliable error recovery. Specialists with such background are few and far between on the market, and the fact that Meta deliberately hunted for the Vercept team only further confirms the strategic value of this direction.
For Anthropic, the acquisition is not merely an addition to the patent portfolio or workforce. The company is consistently building infrastructure for an agentic future: the Model Context Protocol, expanded access to tools, capabilities for long-term planning in Claude. Computer use remains the most complex and simultaneously the most attractive frontier in this work. If an agent can not only answer questions but also execute multi-stage tasks in a real software environment — it ceases to be a chatbot and becomes a full-fledged digital employee. It is precisely this promise that drives the interest of corporate clients willing to pay for automation of routine processes.
The industry as a whole is moving toward a model in which AI agents will not supplement humans with suggestions, but will take on the execution of specific tasks from start to finish. This changes reliability requirements: an agent that makes mistakes when filling out a form or loses context in the middle of a task is not just useless — it is dangerous in a corporate environment. This is why the technologies that Vercept developed — tools for precise, robust interaction with applications — prove to be critically important for any company claiming leadership in the agent segment.
The acquisition of Vercept is Anthropic's conscious bet that the next major round of competition in the AI sphere will be played out not in the space of language models, but in the space of actions. Whoever first teaches their agent to work with the real world reliably and at scale will gain an advantage that will be extremely difficult to catch up with. Judging by everything, Anthropic understands this very well.
Want to stop reading about AI and start using it?
AI News is a curated feed of AI/tech news. Hamidun Academy teaches you to use AI systematically in your work.