Anthropic launched Dispatch today, a new feature inside Claude Cowork that turns your desktop into a persistent AI workstation you can manage from your phone. Scan a QR code, assign a task from anywhere, and come back to finished work. That is the pitch — and it is a meaningful step beyond what most AI tools currently offer.
What Dispatch actually does
Dispatch creates a single synced conversation thread between your desktop app and the Claude mobile app. Your computer does the work; your phone is the remote control.
The flow is simple: open Claude Desktop, scan the QR code on your phone, and you have a live channel to the agent running on your machine. You can assign tasks — pull a report from local files, draft something from your Google Drive, clean up a spreadsheet, handle an email — and Claude executes them on the desktop while you are away from your desk.
Every action requires confirmation. Files stay local. Code runs in a sandbox. Anthropic has been deliberate about the security model here: this is not a system running silently in the background with full access to your machine. It is a supervised agent that needs the app open and the computer awake to function.
Where it fits against the competition
The context matters. OpenClaw — which evolved from Moltbot, covered here — has built a following this year by letting people run an always-on Claude agent on a cheap VPS, accessible via Telegram or Discord. It requires Docker or WSL, a SOUL.md config file, and some tolerance for setup friction. For the technically inclined, it is genuinely powerful.
Dispatch is Anthropic’s answer to that use case without the friction. No VPS, no Node.js, no startup sequence. It works with the files and applications already on your machine — local documents, Slack, connected plugins. The tradeoff is that your computer must be on with the app open. There is no background service, no push notification when a task completes, and no parallel task execution yet.
That is a real constraint. OpenClaw’s value proposition is the 24/7 aspect — tasks that run while you sleep, on infrastructure that stays alive regardless of whether your laptop is open. Dispatch does not replace that. What it does replace is the DIY effort most people were going through to approximate that behavior in a less robust way.
Who it is for right now
Dispatch is rolling out in research preview, starting with Max plan subscribers at $100–200 per month. Pro plan access follows shortly after.
The Max tier qualifier signals that this is early. Anthropic is getting real usage data before a broader rollout, and the feature set reflects that. Single thread, no background execution, no task completion notifications — these are limitations that a research preview can afford, but a polished product would need to address.
For knowledge workers who spend significant time at a single machine and want to hand off tasks during the day, Dispatch is immediately useful. For anything requiring overnight runs or truly autonomous background operation, OpenClaw or a similar always-on setup still has the edge.
The bigger picture
Dispatch is the third major surface in Anthropic’s agent ecosystem, sitting alongside Claude Code (developer-focused, terminal-driven) and the standard Claude.ai interface. Each targets a different user. Dispatch targets the professional who is not an engineer but has real, recurring computer-based work to delegate.
The pattern across the industry is consistent: every major AI lab is working toward AI that takes multi-step actions on your behalf rather than just answering questions. OpenAI shipped native computer use in GPT-5.4 earlier this month; now Anthropic is pushing the same capability into a consumer product surface. Dispatch does not break new technical ground — the underlying computer use capabilities have been public for months — but it packages them into something accessible enough that non-technical users can actually use it.
That is often where the real adoption happens.
For developers and enterprise teams evaluating AI automation tooling, Dispatch is worth watching alongside the broader wave of AI agents rewriting automation workflows. The research preview will reveal where the rough edges are, and the follow-up releases should address background operation and notifications. If Anthropic gets those right, Dispatch becomes a serious competitor to the patchwork of tools people currently use to approximate always-available AI assistance.
Claude Dispatch is available today in research preview for Max plan subscribers. Anthropic’s Cowork documentation covers setup and current limitations.