Anthropic releases computer use so Claude can operate a desktop

A model that clicks, types and reads a screen is a procurement actor, not only a chatbot.

On 22 October 2024 Anthropic put computer use in public beta on Claude 3.5 Sonnet. Developers can direct the model to look at a screen, move a cursor, click and type. Anthropic described the skill as experimental and error-prone. It is not a payment protocol, but it is a precondition for agents that can complete existing human workflows without a dedicated API.

Why it matters

Once a model can operate existing UIs, the commercial question becomes whether those UIs speak a machine payment rail — or whether the agent is stuck scraping checkout pages.

Source: Anthropic

Related

• Anthropic company
• Anthropic machine twin machine
• Model Context Protocol protocol
• Model Context Protocol machine twin machine
• what is an agent wallet explainer
• Give an agent a safe spending budget tutorial
• Indexed machine market market
• Visa, Mastercard and Ant International begin Know-Your-Agent interoperability work prior-story
• Agent wallets converge on stablecoins plus policy limits prior-story
• Visa Intelligent Commerce opens the card network to AI agents prior-story