This photo shows a figurine placed in front of the logo of the AI assistant “Claude” manufactured by American artificial intelligence safety and research company Anthropic during a photo session in Paris on February 13, 2026 (Photo: Joel Saget/AFP via Getty Images)
Joel Saget | AFP | Getty Images
Anthropic’s Claude aims to develop an AI agent that rivals the viral OpenClaw, allowing it to use human computers to complete tasks.
Users will now be able to message Claude with tasks from their phones, and the AI agent will complete the task, Anthropic announced on Monday.
When prompted, Claude can open an app, navigate a web browser or fill out a spreadsheet on his computer, Anthropic said. One of the prompts shown in a video Anthropic posted on Monday is about a user being late for a meeting. The user asks Claude to export the pitch deck as a PDF file and attach it to the meeting invitation. The video shows Claude performing the task.
Anthropic’s latest update highlights the AI company’s progress in creating so-called “agents” that can autonomously perform tasks on your behalf at any time.
This year, with the release of OpenClaw, agent functionality has been in the spotlight. OpenClaw links to AI models from OpenAI and Anthropic. Users can send messages to OpenClaw through popular apps such as WhatsApp and Telegram to perform tasks. As with Anthropic’s new features, OpenClaw runs locally on users’ devices and allows them to access their files.
Nvidia CEO Jensen Huang told CNBC last week that OpenClaw is “definitely the next ChatGPT” as technology companies race to build their own competitors. Last week, the chip leader announced NemoClaw, an enterprise-grade version of OpenClaw.
Safety measures
Anthropic warned that his use of computers is “still early compared to Claude’s coding and text manipulation abilities.”
“Claude can make mistakes, and while we continue to improve our safeguards, threats are constantly evolving,” Antropic warned.
The company has built computer usage features “with safeguards to minimize risk,” Claude said, adding that it always requests permission before accessing new apps.
Users can use Dispatch, a feature released last week in Claude Cowork. This allows users to continuously talk to Claude and assign tasks for the agent from their phone or desktop.
