Claude is Gaining on OpenClaw

Anthropic has been on a tear in adding features to its Claude AI agents for coding and other white-collar work. The Claude agents now offer many of the features that turned OpenClaw, an open-source tool for developing personal AI agents, into a sensation.

Claude’s new features include the ability to take over a person’s computer to perform complex tasks using any application, which Anthropic launched with great fanfare Monday night; take commands from its human users through messaging apps like Telegram and iMessage; and remember details from previous jobs or perform routine tasks on a schedule.

Claude, which only requires a desktop app, is relatively painless to install compared to OpenClaw, which requires entering commands in a Terminal window and configuring it to work with the right models and messaging apps. By default, Claude Cowork also asks permission before editing files and runs in a “virtual machine” separate from a computer’s main operating system, which helps it avoid some of the cybersecurity concerns that have followed OpenClaw. (Those security issues are why some users shell out for Mac Minis or specialized Nvidia computers to keep their OpenClaw agents away from their other devices.)

But OpenClaw is still kicking, and how! Data collected through NPM, software for downloading programs such as OpenClaw, shows that OpenClaw was downloaded more than 400,000 times on Tuesday, near the all-time high of 500,000 daily downloads earlier this month. And those figures don’t include downloads of the countless spinoffs of OpenClaw, such as Nvidia’s NemoClaw.