This {photograph} exhibits a figurine in entrance of the emblem of the AI assistant “Claude” constructed by the US synthetic intelligence security and analysis firm Anthropic throughout a photograph session in Paris on February 13, 2026. (Picture by Joel Saget / AFP through Getty Photos)
Joel Saget | Afp | Getty Photos
Anthropic’s Claude can now use an individual’s pc to finish duties as the corporate appears to create an AI agent that may rival the viral OpenClaw.
Customers can now message Claude a process from a cellphone, and the AI agent will then full that process, Anthropic introduced Monday.
After being prompted, Claude can open apps in your pc, navigate an online browser and fill in spreadsheets, Anthropic mentioned. One immediate Anthropic demonstrated in a video posted Monday is a person operating late for a gathering. The person asks Claude to export a pitch deck as a PDF file and connect it to a gathering invite. The video exhibits Claude finishing up the duty.
The newest replace from Anthropic underscores the push from AI companies to create so-called “brokers” that may autonomously perform duties on behalf of customers at any time of day.
Agentic capabilities have been thrust into the highlight this 12 months after the discharge of OpenClaw, which went viral. OpenClaw hyperlinks to AI fashions from OpenAI and Anthropic. A person can message OpenClaw by fashionable apps like WhatsApp or Telegram to hold out duties. Like Anthropic’s new characteristic, OpenClaw runs domestically on a person’s machine giving it entry to recordsdata.
Nvidia CEO Jensen Huang advised CNBC final week that OpenClaw is “undoubtedly the subsequent ChatGPT” as tech firms race to construct their very own rivals. The chip chief final week introduced NemoClaw, an enterprise-grade model of OpenClaw.
Safeguards
Anthropic cautioned that pc use “remains to be early in comparison with Claude’s potential to code or work together with textual content.”
“Claude could make errors, and whereas we proceed to enhance our safeguards, threats are continuously evolving,” Anthropic warned.
The corporate added that it has constructed the pc use functionality “with safeguards that decrease danger,” and that Claude will at all times request permission earlier than accessing new apps.
Customers can use Dispatch, a characteristic it launched final week in Claude Cowork. That lets customers have a steady dialog with Claude from a cellphone or desktop and assign the agent duties.
