Claude Dispatch Ultimate Guide (How Beginners Can Automate Their Lives)

Welcome to the first iteration of the AI Edge newsletter!

Every week, we cut through the noise and give you what actually matters in AI - the breakthroughs, the tools, and the strategies to ultimately help you become “the AI person.”

This week, we’ll be covering Claude Dispatch. Essentially, Anthropic’s version of OpenClaw.

This is the lowest-hanging fruit for getting started with AI automation, and most people have no clue that it just launched.

You can think of Claude Dispatch as Claudebot.

It is quite literally Anthropic's version of OpenClaw, built directly into the Claude tool suite.

Dispatch lives inside Claude Cowork and lets you text tasks to your desktop AI agent in a single continuous thread via the Claude mobile app.

This continuous thread automatically syncs across your phone and desktop for a seamless task delegation process.

Essentially, Dispatch is a remote-control session of Cowork that gives it access to all your files, context, and tools.

What Dispatch can actually do

If a human could do it, Dispatch can automate it now.

Here is what that actually looks like in practice:

In the last section of this guide, we provide practical Dispatch prompting frameworks to test now

Dispatch setup is genuinely one of the fastest onboarding experiences Anthropic has ever shipped. No API keys, no configuration files, and no OAuth like OpenClaw.

Here is what you need before you start:

Once you have all the requirements, there are four steps you need to follow:

Open the Claude desktop app. In the left sidebar, click Cowork. You will see Dispatch listed as an option; click it.

Within the main Dispatch screen, you'll see "Get Started."

2. Pair your phone

After hitting Get Started, a QR code will appear on your desktop screen.

Open the Claude mobile app on your iPhone, go to the Cowork tab, tap Pair with your desktop, and scan the QR code.

This will connect your two devices.

3. Configure your settings

Once paired, you'll see two features worth enabling:

Give Claude access to your files: this allows Dispatch to pull from your local folders when executing tasks. Turn this on.

Keep your computer awake: this prevents your Mac from sleeping while Dispatch is running a task. Turn this on (if your computer sleeps mid-task, Claude stops working).

4. Send your first task (test)

Open the Dispatch section in your Claude mobile app.

Type a task, and send it.

If properly connected, you should be able to watch Claude execute the task on your desktop.

Done. Your Claude Dispatch is now fully set up and ready to use.

Our team has been testing Dispatch since its release, and these are our Pro Tips (thus far):

Most of this guide has been technical thus far. Now, we’re going to cover some personal opinions on both of these agentic tools.

TLDR: Dispatch wins in simplicity, but OpenClaw wins in customisation.

With Dispatch, you're opting to use Anthropic's tool right out of the box. It's easy to set up, gives you instant access to the entire Anthropic suite (your data, skills, connectors), and just works well right away.

However, in a way, you're also limited to Anthropic's tool suite, and you're relying on them to manage/keep improving the tools over time (for example, Dispatch is only on Mac right now). We have also personally noticed that Dispatch is pretty expensive, especially when executing "real work" because Claude is taking screenshots, analysing the image, thinking, etc.

On the other hand, with OpenClaw, you can run any model you want (this can be significantly cheaper), and you have access to more customisations to build whatever agentic experience you want. Like, texting via WhatsApp or building sub-agents in Discord. You can also run OpenClaw locally, which can offer better privacy.

The obvious downsides of OpenClaw are the setup, maintenance, and security risks.

There are some other minor nuances when it comes to comparing the two:

In summary, if you're planning to build complex workflows, you'll likely need OpenClaw, but there's no harm in setting up Dispatch anyway and using it for light tasks.

Now that you understand what Dispatch is and how it works, here are real prompts that you can test now.

Remember, Dispatch can do any computer task that a human can.

You can use this prompt style with any app or tool.

"Open my Notion and add [x] to my to-do list."

"Send out Google Calendar invites to [[email protected]]."

"Log into my Instagram and give me a summary of my DMs."

The possibilities of connector/app scanning are endless.

This prompting strategy is very powerful. It's like having a personal assistant.

Essentially, you tell Cowork/Dispatch that you're stepping away, and that implies you want the agent to handle the entire project end-to-end.

You can use this for file management, cleaning up your to-do list, clearing reminders, or really any small admin tasks at the end of your workday.

For me personally, just using Dispatch for these types of tasks has been a high ROI.

Tip: When used often, Dispatch can get pretty cluttered. I've personally found it useful to "delete conversation" and "archive" chats.

It makes the Cowork interface a lot cleaner, and it feels like you're starting from scratch.