Anthropic Puts Claude in the Driver's Seat on Your Mac — And It's Aiming at OpenClaw
Anthropic just crossed a significant threshold: Claude Code can now take the wheel on your Mac. As of this week's research preview, Claude Code and Claude Cowork are capable of pointing, clicking, opening files, running dev tools, and navigating browsers autonomously — no MCP server, no external connector, no human in the loop. The rollout is limited to Claude Pro and Max subscribers on macOS for now, but the implications reach well beyond those tiers.
What makes this expansion particularly notable is how Anthropic has designed the fallback model: Claude will first attempt to use purpose-built connectors for services like Google Workspace or Slack. Only when no connector exists does it reach for direct computer control. That architecture keeps the experience predictable for enterprises while still giving developers a powerful escape hatch for anything not yet covered by the MCP ecosystem. Combined with Dispatch — Anthropic's cross-device continuity layer — users can kick off a Mac automation task from their phone and come back to a completed result.
9to5Mac frames this move explicitly as Anthropic entering the computer-use agent arena pioneered by OpenClaw and now contested by Perplexity Computer and Meta's Manus. For Claude Code practitioners, the most practical takeaway is this: the gap between "Claude can help me write this" and "Claude can just do this" just got a lot smaller. Autonomous desktop control as a first-party Anthropic feature — no third-party tooling required — marks the most significant expansion of Claude Code's agentic surface area to date.