In a move that significantly escalates the AI arms race, Anthropic has launched what it calls the most ambitious consumer AI agent to date: Claude computer use, giving its Claude chatbot the ability to directly control a user’s Mac ??clicking buttons, opening applications, typing into fields, and navigating software on the user’s behalf while they step away from their desk.
The feature, now available as a research preview for paying subscribers (Claude Pro at /month and Max at -200/month), represents a fundamental shift in how we interact with AI assistants. This isn’t just a chatbot that answers questions ??this is an AI that can do things for you inside your actual applications.
How Claude Computer Use Works
When a user assigns Claude a task, the system follows a layered priority hierarchy designed to balance reliability with flexibility:
- Direct connectors first ??Claude checks for integrations with services like Gmail, Google Drive, Slack, and Google Calendar. These are the fastest and most reliable paths to task completion.
- Browser navigation second ??If no direct connector exists, Claude falls back to navigating Chrome via Anthropic’s dedicated browser extension.
- Screen-level interaction last ??Only as a last resort does Claude interact directly with the user’s screen ??clicking, typing, scrolling, and opening applications the way a human operator would.
This hierarchy matters enormously in practice. As Anthropic’s documentation explains, pulling messages through a Slack connection takes seconds, but navigating Slack through the screen takes “much longer and is more error-prone.” Screen-level interaction is the most flexible mode ??it can theoretically work with any application ??but it’s also the slowest and most fragile.
What Claude Can Actually Do on Your Mac
When Claude operates at the screen level, it takes periodic screenshots of the user’s desktop to understand what it’s looking at and determine how to navigate. This means Claude can see anything visible on the screen, including personal data, sensitive documents, or private information.
Anthropic has trained Claude with guardrails to avoid:
- Engaging in stock trading
- Inputting sensitive personal data
- Gathering facial images
However, the company is candid: “these guardrails are part of how Claude is trained and instructed, but they aren’t absolute.” This frank admission underscores both the power and the responsibility that comes with AI agents capable of screen-level control.
Dispatch: Your iPhone as a Remote Control for AI
The most strategically significant feature may be Dispatch ??a companion capability that pairs with computer use to create something genuinely novel: the ability to text Claude instructions from your iPhone and watch it execute those instructions on your desktop Mac, even when you’re nowhere near the machine.
A user pairs their mobile device with their Mac by scanning a QR code, and from that point forward, they can assign tasks remotely. Claude executes on the desktop ??which must remain awake and running the Claude app ??and sends back results when complete.
Use cases Anthropic envisions include:
- Morning briefings ??having Claude check your email and prepare a summary while you commute
- Scheduled automation ??”every Friday at 5pm, pull my weekly metrics into a report template”
- Downloads organization ??having Claude tidy a cluttered Downloads folder
- Competitive analysis ??compiling formatted documents from local files and connected tools
The Agentic AI Race Heats Up
The launch thrusts Anthropic into the center of the most heated competition in artificial intelligence: the scramble to build AI agents that can act, not just talk. OpenAI, Google, Nvidia, and a growing swarm of startups are all chasing the same prize ??an AI that operates inside your existing tools rather than beside them.
The stakes are no longer theoretical. Reuters reported that OpenAI is actively courting private equity firms in what was described as an “enterprise turf war with Anthropic” ??a battle in which the ability to ship working agents is becoming the decisive competitive weapon.
One early user on social media captured the broader implications: combining Dispatch with scheduled tasks creates “a background worker that can interact with any app on a cron job. That’s not an AI assistant anymore ??that’s infrastructure.”
Early Performance: It Works About Half the Time
Anthropic is calling this a research preview for good reason. Early hands-on testing suggests the feature works well for information retrieval and summarization tasks but struggles with more complex, multi-step workflows ??particularly those requiring interaction with multiple applications in sequence.
This 50% reliability figure may actually be the point: Anthropic appears to be releasing early to gather real-world feedback and iterate rapidly, rather than waiting for perfection before shipping. Given the pace of AI development, this approach has proven effective for other major releases in the space.
macOS Only for Now, But the Direction Is Clear
Currently, Claude computer use and Dispatch work only on macOS, with Windows and Linux support expected to follow. The research preview designation means Anthropic is explicitly inviting users to help identify failure modes and edge cases before a broader commercial rollout.
The implications for knowledge work, software development, and personal productivity are profound. While the technology isn’t yet at the stage where you can walk away from your desk and trust Claude to handle a complex multi-step project autonomously, the trajectory is unmistakable. The era of AI agents that actually do work ??not just talk about doing work ??has begun.