Claude Cowork Dispatch 101: Remote Control for Desktop AI

Until recently, you could only use personal AI agents while seated in front of a computer. However, that has quickly changed with the release of personal AI assistants that you can chat with on your favorite messaging apps. The agents can then work autonomously on a computer and perform tasks such as reading emails or creating web applications. 

With Claude Cowork, you can use Claude's agentic capabilities on your desktop. This required you to use it on your computer, but with the latest Anthropic update, you can use it on your phone. They are calling it Claude Cowork Dispatch. 

I have been exploring it since it was announced, and in this article, I will share what I discovered. Including how to set it up, use cases, and how it compares to other agents, such as OpenClaw. I also recommend checking out our guides on other Claude remote features, including Claude Code Channels and Claude Code Remote Control. 

TL;DR

• Claude Cowork Dispatch lets you text tasks to your desktop Claude agent from your phone and receive finished results
• Your computer must stay awake and online — Dispatch is a remote control, not cloud computing
• Setup takes under 2 minutes: scan a QR code to pair your phone with Claude Desktop
• Available on Claude Pro ($20/month) and Max ($100-200/month) plans
• Best for information retrieval, file searches, email summaries, and database queries — browser automation is less reliable

What Is Claude Cowork Dispatch?

Claude Cowork Dispatch is Anthropic’s latest update to Claude Cowork, enabling you to use it on your phone. 

You can dispatch tasks from your phone, and the local desktop agent will pick them up and work on them. Claude will run on your computer and message you when the result is ready. This is possible because it can access your local files, connectors, and plugins. 

Claude Cowork Dispatch is powered by a dynamic multi-model network. It uses Claude Opus 4.8 for complex autonomous reasoning, Claude Sonnet 5 for fast everyday task execution, and Claude Fable 5 as an upgrade for elite workloads.

Claude Cowork Dispatch Key Features and Capabilities

Let’s talk about some of the key features and capabilities of Claude Cowork Dispatch.

The feature I found really useful is cross-device persistence. 

For instance, you can message Claude from your phone and later continue the conversation on the desktop. This is possible because it creates a single persistent thread that retains context from previous tasks. 

Next, let’s look at local execution. When you send a message from your phone, it triggers actions on the host device. Claude will be able to perform the task using your existing files, connectors, and plugins. When it's done, it will message you with the asset, for example, a spreadsheet. 

Claude Cowork Dispatch also has unified connector access. It inherits all Desktop MCP connectors, including local files, Notion, Slack, and Google Drive if connected. This means that anything that you could do on the desktop app can now also be triggered from your phone. 

Finally, let’s talk about browser agency. Even on your phone, you will have full control over the host machine. For example, you can have Claude access your Gmail and draft some replies for you to review.  

Setting Up Claude Cowork Dispatch

Let’s look at how to set up Claude Cowork Dispatch. 

Pairing & Permissions

To use Claude Cowork Dispatch, you will need to be on the latest version of Claude Desktop. You will see the Dispatch Tab on the left sidebar. Click it to get started. Scan the QR Code to begin the process. Make sure you have the Claude app installed on your phone. 

You then need to enable keep desktop awake to prevent the host machine from sleeping during remote tasks. 

You also need to grant some permissions, such as access to local directories. 

Using Claude Cowork Dispatch’s Standout Capabilities

Claude Cowork Dispatch has four standout capabilities that separate it from a standard desktop chatbot. 

The persistent thread logic

Claude Cowork Dispatch's persistent thread makes it possible to easily switch between the desktop and mobile app. You can kick off a task while on a walk and pick it up again when you settle down at your computer. 

Local file orchestration

Claude Cowork Dispatch enables you to access the files on your computer from your phone. For example, you can ask Claude to analyze a CSV file from your downloads folder without the data leaving the local sandbox. 

Remote browser control

With remote browser control, you can have Claude access your browser tabs with your authenticated sessions. This means that you can ask Claude to complete a task that you started on the computer by chatting with it on the phone. 

Claude Cowork Dispatch in Action - Real Use Cases

I tested Claude Cowork Dispatch on three workflows: browser automation, local code review, and email drafting. These are inspired by use cases I saw others share after launch. 

Browser automation

First, I asked Claude on my phone to go to YouTube and return the 20 most recent titles for videos covering Claude Cowork Dispatch.

As you can see from the screenshot below, Claude Cowork Dispatch started by asking for permission to access my browser. 

Next, Claude Cowork Dispatch gave me the result: 

Then I wanted to see if it could actually analyze the videos and send me some of the questions people are asking. So I sent the message below from my phone. 

Claude Cowork Dispatch opens all the videos, scrolls to the comments section, and provides the report.

To learn more about browser automation, check out our article on Chrome DevTools MCP. 

Code review

When I tried to get Dispatch to review a local repository and suggest some improvements, the first thing it did was to ask for permissions to access that folder:

It then sent the report on the mobile app and the desktop app, as shown below: 

When working on a real project, it’s better to first create a plan to guarantee good results. You can learn how to use Claude Code’s Plan Mode to explore your codebase in read-only mode, surface questions, and review structured plans for multi-file changes in our Claude Code Plan Mode tutorial. 

Email drafting

Okay, one final example. Let’s see if Claude Cowork Dispatch can draft a reply for the latest email from my editor, Matt. Claude realized that I had no Gmail connectors, asked for permission to access Gmail, found the email, and finally drafted a response as shown below. 

Tips for Getting Better Results with Cowork Dispatch

Based on my testing and community feedback, these practices improve Dispatch's reliability:

• Keep tasks information-focused. Summarize, find, list, retrieve, and compile operations work consistently. Cross-app actions and system-level commands are less reliable.

• Be specific about file locations. "Check /projects/Q1/budget.xlsx" works better than "find my budget somewhere." The less guessing Dispatch needs to do, the more consistent the results.

• Test connectors at your desk first. If Gmail or Notion works in regular Cowork, it will work through Dispatch. Do not try setting up new connectors remotely.

• Prevent your Mac from sleeping. Go to System Settings > Energy and adjust your sleep timer. Dispatch stops the moment your computer enters sleep mode.

• Avoid chaining multi-step tasks. Single-purpose requests like "summarize my emails" succeed more often than compound instructions like "summarize my emails, then create a Notion page from the summary, then share it with my team."

Claude Cowork Dispatch Limitations and Caveats

Claude Cowork Dispatch has four limitations worth knowing before you rely on it for daily work. 

The first one is the host awake requirement. You have to keep your computer online and active for you to send messages from the mobile app. It’s almost like running your computer like a cloud-based virtual machine, which is always on. 

Second, Claude Cowork Dispatch lacks push notifications on mobile. You will not get a notification on desktop or mobile when a task is complete. 

Third, Claude Cowork Dispatch operates on a single thread of conversation. At the moment, there is no way for you to start a new thread or manage multiple threads. 

You also can't set up scheduled or recurring tasks through the Dispatch mobile interface. To create automated schedules (like a daily email summary at 9 AM), you need to configure them directly in Cowork on your desktop.

Privacy and Data Handling

One design choice worth highlighting: Dispatch processes everything locally on your machine. Your phone sends text instructions to your desktop, and your desktop executes them. Files never leave your computer, and data does not route through external servers for processing.

The permission gates from regular Cowork carry over. If a task requires deleting files or modifying documents, Dispatch sends an approval request to your phone before executing. Any global instructions you have set in Cowork (like "never delete files without confirmation") also apply to Dispatch sessions.

For teams with compliance requirements, the local-only architecture means no additional data residency concerns. The trade-off is the host-awake requirement — no always-on cloud server also means no always-on attack surface.

Claude Cowork Dispatch vs OpenClaw

When researching this topic, I saw a lot of people on the internet call Claude Cowork Dispatch the OpenClaw killer. Let’s examine if that is true. 

Security and architecture: Sandboxed vs full system access

By default, OpenClaw runs with full system access unless you set it up with Docker and limit the permissions. This allows for more flexible automation at the risk of the agent doing something bad, like deleting your inbox, as we saw in the OpenClaw vs Claude Code article. 

On the other hand, Claude Cowork Dispatch will ask for permissions when it wants to execute certain commands, such as accessing your directories. This approach makes Claude Cowork Dispatch safer for businesses and enterprise users. 

Ease of use: One-click vs terminal-heavy

Claude Cowork Dispatch is significantly easier and faster to set up than OpenClaw. It only involves downloading the desktop app and scanning a QR code on your phone. It works by default if you are already logged in on your phone and computer. 

OpenClaw is more complex to set up. It requires manual installation, some familiarity with the command line, Docker, and environment variables. You are also responsible for figuring out the right level of permissions and how to deal with network security. 

You can learn how to set up OpenClaw and connect it to WhatsApp in our OpenClaw (Clawdbot) Tutorial. 

Persistence: Session-based vs always-on daemon

Claude Cowork Dispatch works only when your computer is awake and connected to the internet. OpenClaw is an always-on assistant that can monitor its environment, run cron jobs, and execute tasks with little human intervention.

Cost and ecosystem: The subscription vs The API

Claude Cowork Dispatch is a closed application, requiring a Claude Pro or Max subscription. This means that you are locked into Anthropic’s ecosystem. 

OpenClaw is an open-source application that you can self-host and use with any model provider, including open-source models. However, for closed models, you have to pay for each token, unlike in Claude Cowork Dispatch, where model usage is covered in your subscription. 

Discover how to use OpenClaw with Ollama in our OpenClaw with Ollama tutorial. 

Comparison Table

Future Outlook

I would say not many people are surprised that Anthropic launched something similar to OpenClaw. Since OpenClaw has become the most popular software in the world, every company is now scrambling to offer something similar while fixing the main challenge that OpenClaw has: security. 

It's no surprise that Anthropic has launched a similar offering, but with tighter permissions and security. Other model providers, such as Perplexity, have also launched similar tools. 

The Perplexity Computer is basically a hosted version of OpenClaw, allowing you to run persistent tasks and connect to your favorite apps, such as Gmail, Slack, Notion, and even your calendar. 

Claude Cowork Dispatch is Anthropic’s answer to OpenClaw, targeted at enterprises and customers who are security-conscious. It’s also for users who want a personal AI assistant, but don't want to go through the long route of setting up everything from scratch. 

I would expect that Anthropic will ship new features that directly compete with OpenClaw. The one I can bet will be released very soon is the ability to run Claude Cowork Dispatch without your computer being on. 

For example, they could make it possible to dispatch some tasks to the cloud if your computer is offline. I would imagine a task like creating a Notion page should be possible via the API; they would just need to provide a secure way to store the API keys and share them across the mobile and desktop applications. 

Final Thoughts

With Claude Cowork Dispatch, Anthropic proves that a desktop AI agent can run with tight permission controls and still be useful. If you work with files, email, and browser tabs throughout the day, Dispatch removes the requirement to be physically at your computer for every task. 

For a deeper dive into the full Cowork platform, check out our Claude Cowork tutorial, and explore how other desktop agents compare in our guide to the best AI agents in 2026. To start building with Claude's API directly, I recommend our Introduction to Claude Models course.