OpenClaw has become one of the most talked-about AI tools ever, and Anthropic has looked to bring some of its functionality with its own ecosystem.
The San Francisco-based AI company announced that Claude can now use your computer to complete tasks — opening apps, navigating the browser, filling spreadsheets, and doing whatever else you’d normally handle yourself at your desk. Paired with a new feature called Dispatch, users can assign those tasks from their phone and walk away. It is currently in research preview in Claude Cowork and Claude Code, and is limited to macOS.

What OpenClaw Did First
OpenClaw started as a weekend project by Austrian developer Peter Steinberger in November 2025. You could text it on Telegram or WhatsApp and it would manage your calendar, triage your email, run scripts, and browse the web. The formula was simple but the timing was exactly right: it gained 25,000 GitHub stars in a single day and surpassed React’s star count within two months — a milestone that took React over a decade.
Steinberger had connected a messaging API to Claude’s API, added local system access, and had a working prototype in an hour. The irony is hard to miss: OpenClaw was built on top of Claude’s API, yet it was the open-source project that cracked the agentic interface problem first. As Steinberger put it, big companies couldn’t do it — not for technical reasons, but organizational ones.
OpenClaw nailed three things previous agent projects missed: proximity, creativity, and extensibility. It lives where users already are — WhatsApp, Slack, Discord, Telegram, Signal. That single design decision changed its trajectory.
Anthropic’s Answer: Computer Use + Dispatch
Anthropic’s response is more tightly integrated into its own product stack. Claude can now directly control your browser, mouse, keyboard, and screen when it lacks a dedicated connector to a service. It will scroll, click, and navigate as needed, always asking for permission before accessing new applications.
The company says it has built in safeguards against prompt injection — when Claude uses your computer, its system automatically scans model activations to detect such activity. Users can stop Claude at any point, and certain sensitive apps are off-limits by default.
Paired with computer use is Dispatch — a feature that enables one continuous conversation with Claude across phone and desktop. Assign a task from your phone, switch your attention elsewhere, and return to finished work on your computer. You can tell Claude to check your email every morning or generate a report every Friday, and it handles scheduling from there. Dispatch is available in both Claude Cowork and, as of this announcement, Claude Code.
The combination opens up meaningful workflows: drafting a morning briefing while you’re on the commute, making changes in an IDE and pushing a pull request, or keeping a long-running project progressing according to an initial plan — all without sitting at a computer.
The Honest Caveats
Anthropic isn’t overselling this. The company acknowledges that computer use is early-stage relative to Claude’s text and coding capabilities, that complex tasks may need a second attempt, and that working through the screen is slower than a direct integration. Security threats are also a stated concern — the company recommends starting with trusted apps and avoiding sensitive data.
This measured approach is a pointed contrast with OpenClaw’s origins. A security audit of OpenClaw in late January 2026 identified 512 vulnerabilities, eight of which were classified as critical. When a cybersecurity firm flagged the issues, Steinberger’s response was blunt: it was a hobby project, not a production system. There have also been reports of agents deleting entire email inboxes during automated cleanup workflows.
Anthropic has enterprise obligations and liability considerations OpenClaw never had to worry about. The tradeoff is a more guarded rollout, but also one that organizations can plausibly consider deploying.
Who Gets It
Computer use and Dispatch are available to Claude Pro and Max subscribers. The desktop app must be running for computer use to function, after which it can be paired with the mobile app to hand off tasks remotely. For now, macOS is the only supported platform.
Given Claude’s trajectory on agentic benchmarks — Opus 4.6 scored 72.7% on OSWorld, the standard computer use benchmark — the underlying model is capable. The question is whether Anthropic can make the experience seamless enough to compete with an open-source tool that has, at last count, 250,000 GitHub stars and a devoted community that treats it like early AGI.