While everyone was still waiting for OpenAI to fully unleash its long-teased computer-control agent (affectionately dubbed “OpenClaw” in the community), Anthropic quietly dropped the feature first.
Say hello to Dispatch inside Claude Cowork — a new research preview that lets you assign tasks to Claude from your phone and have the AI run them on your desktop while you’re away.
How It Actually Works (and Why It Feels Like Magic)
You keep the Claude Desktop app open on your Mac (or Windows machine). It must stay awake and connected.
You open the Claude mobile app on your phone, jump into the single persistent Cowork thread, and simply tell Claude what you need: “Summarize my inbox, pull the latest sales numbers from Notion, create a presentation, and export it as a PDF.”
Claude on the desktop goes to work exactly as it always does in Cowork:
- Reads your local files;
- Uses all your configured connectors and plugins;
- Browses, analyzes, builds spreadsheets, writes reports.
The best part? You don’t see the intermediate steps. No live screen-sharing, no scrolling through 47 tool calls. When it’s done, the final result — clean table, document, presentation, whatever — just appears in the same chat thread, automatically synced to your phone.
One continuous conversation that never resets. Full context. Everything works exactly like the desktop version you’re already used to.
The Fine Print (Because It’s Still a Preview)
- Currently rolling out first to Claude Max subscribers, then Pro users
- Only one persistent thread for now
- Desktop must stay on and the app open
- Claude won’t message you proactively — it only responds when you ask
- Available on Mac and Windows desktop + iOS/Android mobile
Security Reality Check (Anthropic Is Honest About It)
The company itself warns: giving your phone remote control over a desktop AI agent that can read, move, or delete files and interact with services is powerful… but also risky.
Mistakes or bad instructions can have real consequences.
Everything runs in your own environment with your own connectors, which feels safer than cloud-only agents, but the usual “don’t give AI the nuclear codes” advice still applies.
Also read:
- AI-Oscar 2026: QUASA Readers Crown Their Favorite AI Tools in Historic People’s Choice Awards
- SEC Makes History: First Official Definition of Crypto Securities Released – Most Digital Assets Declared Non-Securities
- The Not-So-Obvious Truth: OpenClaw's Rise Is a Direct Consequence of DeepSeek's Breakthrough
Why This Matters
This isn’t just another gimmick. It’s the first real taste of the future everyone has been predicting: your AI actually doing desk work while you’re on the train, in a meeting, or at the gym. No need to babysit the agent. You come back to finished deliverables.
And the fact that Anthropic shipped this version before OpenAI finished polishing its own computer-use agent has the entire community buzzing. “Anthropic built OpenClaw faster than OpenAI” is the meme of the day — and it’s not wrong.
Dispatch is still early (research preview, one thread, no auto-pings), but the direction is crystal clear. Every major AI vendor is heading exactly here: true remote AI employees you can boss around from anywhere.
If you have Claude Pro or Max, go grab the latest desktop and mobile apps and try it. Your new remote coworker is already waiting on your desk — and you can give it orders from the beach.
