Claude Dispatch: What It Is, How It Works, and What to Know

7 min read

Most professionals learned about Claude Cowork as a desktop agent — a local-first tool that mounts your folders, edits your spreadsheets, and runs inside a VM on your machine. Dispatch extends that model to your phone. You can now text Claude a task from anywhere, and it executes the work on your desktop while you're away from it.

Launched as a research preview on March 17, 2026, Dispatch is available to Claude Pro and Max subscribers. It is essentially a persistent, two-way thread between your mobile device and your desktop agent. The concept is simple: your phone becomes the delegation layer, and your computer remains the execution layer.

Here's what you need to know.

What is Dispatch?

Dispatch is a feature within Claude Cowork that creates a single, continuous conversation thread synced between the Claude mobile app (iOS/Android) and the Claude Desktop app (macOS/Windows). You message Claude from your phone the same way you'd Slack a junior analyst — describe the task, include any context, and walk away. Claude figures out whether the work requires a Cowork session or a Claude Code session, spins up the right environment on your desktop, and messages you the result when it's done.

The key distinction from regular Cowork: you don't need to be sitting at your computer. The work still runs locally on your machine, but the instruction and the output travel through your phone.

What do I need to set it up?

Four things:

A Claude Pro or Max subscription.
The latest version of the Claude Desktop app, installed and running on your computer (macOS or Windows x64).
The latest version of the Claude mobile app on your phone.
An active internet connection on both devices.

Setup takes about two minutes. Open Cowork on either device, navigate to the Dispatch tab in the sidebar, walk through the onboarding screens, and toggle on file access and the "keep computer awake" setting. Once paired, the thread syncs automatically across both surfaces.

What can I actually do with it?

Anything you could do in Cowork, but initiated from your phone. That includes tasks that rely on local files, desktop apps, connectors, and plugins you've already configured. Some practical examples:

Ask Claude to pull data from a local spreadsheet and compile a summary.
Have Claude scan your Slack messages and email, then draft a briefing document.
Request a formatted presentation built from files in your Google Drive.
Tell Claude to organize or process files in a specific folder on your machine.
Set up recurring tasks — a morning email digest, a weekly metrics pull, a Friday summary report — and Claude handles them on schedule without being asked again.

When Claude finishes a task that produces a file, you can access it directly from the mobile thread or find it on your desktop at the location Claude specifies.

Does Claude remember context between tasks?

Yes. Dispatch uses a single persistent thread. It doesn't reset between tasks. Claude retains context from everything you've worked on previously — your preferences, your projects, your file structures. You don't re-explain things every time you open the app.

You control what Claude remembers. Memory can be viewed, edited, and deleted at any time.

What about computer use?

Dispatch supports Claude's computer use capability, which was integrated shortly after launch. If you ask Claude to update a spreadsheet in Excel, navigate an internal dashboard, or interact with a desktop application that doesn't have a direct API connector, Claude can operate those apps directly on your screen.

This is powerful but worth understanding clearly: computer use through Dispatch runs outside the Cowork sandbox. Claude is clicking around your actual desktop, not a virtualized environment. The safety implications are real and worth reading Anthropic's guidance on before enabling.

What are the current limitations?

Dispatch is a research preview, and it shows in a few places:

Your desktop must stay awake. If your laptop lid is closed, your machine sleeps, or the Claude Desktop app isn't running, tasks pause. There's no cloud fallback. The execution layer is your physical computer.
Computer use runs outside the sandbox. When Claude operates your desktop apps, it's working on your actual machine, not inside Cowork's protected VM. Mistakes or unexpected behavior can have real consequences on your files and connected services.
One thread only. There's no way to create multiple threads or manage separate conversation streams. Everything lives in a single continuous conversation.
Desktop dependency is absolute. This isn't a cloud agent. If your computer is off, Claude is off. This makes Dispatch less useful for professionals who work primarily from a laptop they close between sessions.

Is it safe?

Anthropic is transparent about the risk model here, which is worth acknowledging. Dispatch creates a chain where instructions from your phone trigger real actions on your computer — reading, moving, or deleting files, interacting with connected services, controlling your browser, and operating desktop apps. A manipulated instruction or a phishing link encountered during browsing could cascade into actions that are difficult to undo.

Before enabling Dispatch, you should understand what files and accounts are accessible, trust every app and service in the chain, and know how to quickly disconnect or revoke access. This is not a theoretical concern — it's the practical reality of giving a mobile AI agent remote control of a desktop AI agent.

Who is this actually for?

Dispatch is most useful for professionals who spend significant time away from their desk but still need work produced on their desktop. The commute-to-meeting-to-lunch cycle where you're generating tasks mentally but can't execute them until you're back at your machine — that's the gap Dispatch fills.

For deal teams specifically, the use cases are straightforward: kick off a model refresh from the cab, ask for a document summary while waiting for a meeting to start, have Claude organize a data room folder while you're on a call. The value is in the async delegation, not in any new capability. Everything Dispatch does, Cowork already did. Dispatch just lets you trigger it from your pocket.

Where Dispatch Falls Short for Deal Teams

The honest answer is that Dispatch inherits all of Cowork's strengths and all of its gaps. It's still a general-purpose agent. It doesn't inherently know how your firm formats an IC memo, what columns belong in a rent roll reconciliation, or how to structure a sensitivity table with 25bps increments. Every time you assign a task from your phone, you're still teaching Claude the specifics of your deliverable from scratch — just from a smaller screen.

The delegation model is compelling. But delegation only works well when the person (or agent) you're delegating to already understands the job.

Purpose-built AI coworkers — like those from Lumetric — are designed around that principle. They combine the raw power of frontier models with native understanding of specific industries and deliverables: lease abstracts, CAM reconciliations, deal screens, lender packages. And the "delegate from anywhere" problem that Dispatch is trying to solve with a paired mobile app? Lumetric's coworkers are already reachable by email — the interface your team uses a hundred times a day, from any device, without a new app or a desktop that has to stay awake. Send a task the same way you'd delegate to a junior analyst, and the output comes back in the format your team expects. No prompt engineering, no re-teaching, no desktop dependency.