Anthropic unveils AI agent Claude to complete tasks on your computer

Claude, developed by Anthropic, is now capable of utilising a person's computer to perform tasks as part of the company's efforts to introduce an AI agent that could rival the popular OpenClaw.

As of Tuesday, users can now instruct Claude via phone to execute tasks, which the AI agent will complete, Anthropic revealed.

Upon receiving directions, Claude is capable of launching applications on your computer, surfing the web, and managing spreadsheets, according to Anthropic. 

A situation demonstrated in a video shared Monday shows a user running late to a meeting requesting Claude to convert a pitch presentation into a PDF and attach it to a meeting invite. The video depicts Claude successfully completing the task.

This latest enhancement from Anthropic highlights the trend among AI companies to develop "agents" capable of independently performing tasks for users around the clock.

The conversation around agent capabilities saw significant attention this year following the debut of OpenClaw, which gained widespread attention. OpenClaw connects with AI models from OpenAI and Anthropic. 

Users can communicate with OpenClaw through widely-used apps like WhatsApp or Telegram to execute tasks. 

Similar to Anthropic's new capability, OpenClaw functions directly on users' devices, granting access to files.

Nvidia CEO Jensen Huang mentioned to CNBC last week that OpenClaw is poised to become "the next ChatGPT" as technology companies rush to create their versions.

The leading chip manufacturer announced NemoClaw last week, a professional version of OpenClaw aimed at enterprises.

OpenAI recently recruited Peter Steinberger, the founder of OpenClaw, as part of its mission to "lead the development of the next wave of personal assistants."

Anthropic noted that using computers is "still developing compared to Claude's proficiency in coding or text interaction."

"There is a risk of errors by Claude, and while we are continuously enhancing our security measures, new threats arise all the time," Anthropic advised.

The company emphasised that they have implemented this computer-use feature "with measures to reduce risks" and stated that Claude will always seek approval before engaging with new applications.

This new feature can be utilised through Dispatch, introduced last week in Claude Cowork. 

This allows users to engage in ongoing conversations with Claude via phone or desktop and assign tasks to the agent.