Claude’s AI Just Killed OpenClaw

Claude's AI just killed OpenCLAW. So two downloads, one QR code, your AI agent is live. That's it. That's what Anthropic just did to OpenCLAW.

I need you to sit with that for a second because for the last two months the entire AI world has been losing its mind over OpenCLAW. 300,000 GitHub stars, the fastest growing open source project in history, people buying dedicated Mac minis, setting up Node.js servers, configuring WebSocket gateway, spending entire weekends just trying to get the AI agent to respond to a text message on Telegram. And on March 17th, 2026, Anthropic dropped a feature called Dispatch. And it does what OpenCLAW does except it takes two minutes to set up instead of two days.

No servers, no command line, no API keys to configure, you download Claude desktop, you download the Claude app on your phone, you scan the QR code, done. You're texting your AI agent from your phone whilst it works on your computer. It's not a small thing. That is a category killer.

Let me tell you why this matters more than most people think. OpenCLAW was a dream. I get it. I've talked about OpenCLAW on this channel a bunch of times.

The idea is incredible. You have an AI agent that runs 24-7 on your machine. It reads your emails, manages your calendar. The problem was never the idea.

The problem was getting there. Let me paint you in a picture of what setting up OpenCLAW actually looks like if you're a normal person. First, you need Node.js version 2022 or higher stored on your machine. 22, not 2022.

If you don't know what Node.js is, you're already in trouble. Then you need to clone a GitHub repository. Then you need to install dependencies with either npm or pmpm. Then you need to configure your API keys for whatever AI model you need.

Then you need to set up a WebSocket gateway. From there, you connect your messaging platform. If you want to use Telegram, for example, you need a bot token from Botfather. WhatsApp, you need the business API or a third party bridge called Baileys.

If you want to use iMessage, you need a separate tool called Blue Bowls. And that's before you've actually taught your AI agent anything. That's just getting it to turn up. So one of OpenCLAW's own maintainers literally said, and I'm not making this up, if you can't understand how to run a command line, which 99.9% of people can't, this project is too dangerous for you to use.

And that's coming from the people who built it. Now, look, I'm not dunking on OpenCLAW to be mean. It's an amazing tool. It's impressive.

They've done something absolutely incredible with it, right? It's a real project with real power behind it. But here's what I keep saying on this channel. The best technology doesn't always win.

The easiest technology wins every single time. And that's exactly what Anthropy just figured out. Here's how Dispatch works. You open up CLAW Desktop on your Mac or PC.

You click the new Dispatch option in Cowork. It shows you a QR code. You scan with the CLAWed app on your phone. And that's it.

You're paired. One persistent conversation that runs on your computer and shows up on your phone. Then you can text it from your phone on the train. It processes the work on your desktop.

You come back to finish work. Felix Reisberg, who works at Anthropy, can announce this thing, put it perfectly. One persistent conversation with CLAWed that runs on your computer. Message it from your phone.

Come back to finish work. No Node.js, no WebSocket gateway, no 48 startup sequence, no dedicated server running in your closet. Scan and go. Now I can hear the OpenCLAW fans already.

But Julian, OpenCLAW can connect to 20 different messaging platforms. It works with WhatsApp, Telegram, Slack, Discord, Signal, iMessage, even Twitch. And that's true. OpenCLAW's multi-channel support is genuinely impressive.

Dispatch only works through the CLAWed mobile app right now. That's a real limitation. I also hear the people saying OpenCLAW is free and open source. You can use any AR model you want.

CLAWed, GPT, Gemini, even local models through Allama. Also true. Dispatch requires a CLAWed subscription. MaxPlanet, $100, gets it right now.

Pro, $20 a month, gets it in a few days. And I hear the power users saying OpenCLAW runs as a background demon, right? Dispatch needs your computer to be awake with CLAWed desktop open. Close the lid, it stops.

And they're all valid points. I'm not going to pretend otherwise. But here's the thing that nobody's talking about. None of those advantages matter if people can't get Bast to set up, right?

This is the exact same pattern we've seen outplay a hundred times in tech history. The open source project does it fast. The big company does it easier. And 90% of the market goes with easier.

Right? Think about it. Linux was first, but most people use Windows or Mac. WordPress could do anything if you knew PHP, but Squarespace won the small business market because you could drag and drop.

Self-hosting your email gave you total control, but Gmail gave you convenience. And Gmail won. OpenCLAW is Linux. Dispatch is Gmail.

And before anyone else says I'm oversimplifying, let me give you some numbers from the real world. Peter Svodjak, a Czech tech writer, wrote about this comparison the day Dispatch launched. He pointed out that he watched people spend hours configuring OpenCLAW, Docker containers, environmental variables, debugging cryptic Node.js errors about missing import statements because they had the wrong version installed, and then he set up Dispatch, two downloads and a QR code done in five minutes. His exact framing was that Dispatch does what people have been trying to do with OpenCLAW just for the pain.

Mac Stories, one of the most respected Apple publications, tested Dispatch and found it worked reliably for file lookups and information retrieval tasks. They said it's about a 50-50 success rate on more complex stuff. That's honest. It's a research preview.

It's early. But the 50% that works, it works immediately. No configuration required. And the FindSkillAI team ran real tests, finding files on a Mac by searching file names and metadata, quick and reliable, summarizing Notion nodes through the Notion connector.

Worked smoothly. They sent tasks from their phone and got clean results back. Is it perfect? No.

Is it ready to replace a fully configured OpenCLAW setup for power users? Not yet. But that's not the point. The point is that for the first time, a normal person can have an AI agent working on their computer whilst they're standing in line at the grocery store.

And they didn't need to learn what a WebSocket is to get there. Now, let me tell you what has me paying attention. This isn't just about Dispatch. This is about the full stack that Anthropic has been building since January.

And when you see all the pieces together, you start to understand why Simon Willison and Ethan Mollick are comparing it favorably to OpenCLAW. Why the Latent Space podcast, for example, called it Anthropic's answer to OpenCLAW. Here's a stack. Claude Cowork launched in 2026.

It brought Claude Code's agent abilities to non-coders. It can run locally on your computer in an isolated virtual machine. It can read, edit, and create files. It can browse the web.

It can connect to external services. Anthropic launched connectors to Google Drive, Gmail, DocuSign, Apollo, Clay, Outreach, and more. Then they added plugins. Customizable, shareable automations for specific tasks, right?

They added schedule and recurring tasks. So you can just tell Claude to do something early morning or every morning or every Friday afternoon, whenever you need. And they added a plugin marketplace with admin controls for teams and enterprises. And now you've got Dispatch, the remote control, the mobile interface, the thing that cuts the cord between you and your desk.

Each piece on its own is interesting. Together, they're building something that looks a whole lot like what OpenCLAW promised, except it comes from a $380 billion company with SOC 2 compliance, HIPAA eligibility, audit logging, and zero data retention options. That compliance piece is not sexy. I know.

Nobody makes YouTube videos about audit trails, but if you run a business, if you handle client data, if you work in healthcare or finance or legal, that compliance piece is the entire ballgame. OpenCLAW is MIT licensed. That means you can do whatever you want with it, but it also means you're responsible for your own security. And security researchers have found over 135,000 OpenCLAW instances sitting wide open on the internet.

That's not a knock on the OpenCLAW team. Security is super hard, but it's a real thing that real businesses have to think about. And Microsoft noticed too. Microsoft, the biggest software on earth, took one look at what was Anthropic building and said, we need that.

They literally licensed CLAW co-work and built CLAW co-pilot co-work on top of it. Same AR model, same agentic framework, but wrapped in Microsoft's cloud infrastructure and enterprise data protections. Now here's where this gets personal for you. I don't care if you're a developer or a teacher, a real estate agent or a freelance designer, what just happened this week is that the barrier to having an AI agent went from weekend project for technical people to two minutes for anyone with a phone.

And the people who figure out what to do with this first are going to have an insane advantage. Think about what dispatch can actually do right now. You're on the train to work. You text your AI agent, scan my inbox, summarize everything urgent, prepare an agenda for my 10am meeting.

And by the time you sit down at your desk, it's done. You're at lunch and someone asks about project status. You text your agent, find the Q1 budget PDF I downloaded last week and give me the top three takeaways. It searches your files and sends back to the answer, right?

You're lying in bed Sunday night and you remember you need an expense report. You text your agent, process all the receipt images in my receipts folder and create a categorized spreadsheet. You wake up Monday morning and it's sitting in your outputs folder. That is what this technology does.

Right now, today, not in some theoretical future. And here's what I think most people are missing. This is happening so fast now. In January, Cowork launched and software stocks dropped hundreds of billions of dollars and Propic's own valuation hit $380 billion.

Microsoft shared $220 billion in market cap in a single week and the iShare's expanded tech software sector ETF dropped almost 5% in one day just on the announcement of Claude Cowork connectors and plugins. That was two months ago. And since then they've added plugins, connectors, scheduled tasks, a marketplace, admin controls, and now dispatch. In two months.

And here's the part that should really get your attention. This is core code we're talking about underneath all of this. The same core code that just got upgraded to 64,000 output tokens by default for Opus 4.6. The same core code that just added voice mode with Pitch to Tor.

The same core code that added the loop command for recurring monitoring tasks. The same core code that got a million token context window. The engine keeps getting more powerful and the interface keeps getting easier. That's a combination that changes everything.

I know there are people watching this who feel overwhelmed, who, you know, keep hearing about new features, new tools, new launches every week, you know, and you're thinking, I can't keep up. I don't even know where to start. And if that's you, you need to join the AI Profit Boardroom. Seriously, this is exactly what it's for.

I break down every single one of these developments, not just what they are, but what they mean for you, how to actually use them in your business, how to make money with them, how to stay ahead of the curve instead of constantly feeling like you're falling behind. Go to the AIProfitBoarding.com. The people in that community are already using this stuff. They're not watching from the sidelines.

They're building real businesses around them. And the gap between the people who understand this technology and the people who don't is getting wider every single day. Now, let me be real about something. I don't think OpenCore is dead.

That title is dramatic, I know. But I think OpenCore's role has just changed dramatically. Here's what I mean. Before Dispatch, OpenCore was the only game in town for non-developers who wanted an AI agent.

It was hard to set up, but it was the only option. So people suffered through the Node.js configuration and the WebSocket debugging because the payoff was worth it. Now there's an alternative that takes two minutes and it comes from the company that makes the best AI model in the world. And it has enterprise-grade security, and it's backed by a company with $380 billion behind it.

So who still needs OpenCore? Power users, developers, people who want to run multiple agents with different AI models, people who need 24-7 always-on operations, people who want to connect to different messaging platforms, people who want total customization and control. OpenCore becomes the Linux server in the basement. Incredibly powerful for the people who know how to use it, but most people are going to use Dispatch because it just works.

And here's the other thing. Claude Code also has something called the Dispatch skill, built by a developer named Basim Eldaf. It's a different thing than Cowork Dispatch. This one runs inside Claude Code itself, unless you spawn background AI workers.

They each get their own fresh context window. You can mix models per task, use Claude for deep reasoning, use GPT for broad generation, use Gemini for speed. The main session becomes a lightweight orchestra, orchestrator, and the actual fans out to the work, fans out to parallel agents. That's a developer version of the same idea.

Your main session stays lean, workers execute in parallel with fresh context, and questions surface to when you need it, right? No polling, no context locks. So now you've got Cowork Dispatch for normal people and for Code Dispatch for developers, two on-ramps to the same future, an AI agent that works whilst you don't. So let me give you a bigger picture of what's happening here.

We're watching the AI industry go from the exact same transition that happened with the internet in the late 1990s. In 1995, if you wanted a website, you needed to know HTML, you needed to rent server space, you needed to configure DNS records. It was a technical person's game. Then GeoCities came along and said, just pick a template.

Then Blogger came along and said, just write. Then WordPress came along and said, anyone can do this. And then Squarespace came along and said, you don't even need to think about it. Each step made the technology more accessible to more people, and each step created a massive new wave of economic value.

OpenClaw is the no HTML era of AI agents. Dispatch is the just write era. And we're not even close to the Squarespace era yet. We're maybe 18 months from agents that configure themselves, learn your preferences automatically, and proactively handle things before you even ask.

But the trajectory is obvious if you're paying attention. But here's what gets me excited. And honestly, what keeps me up at night a little bit too. We're at the point where the conversation is shifting from, can AI do this?

To how fast can AI learn to do this? And the answer keeps getting faster. In January, Cowork launched and could work with local files. By February, it had connectors to Google Drive, Gmail, and DocuSign.

By March, it has Dispatch, Scheduled Tasks, a plugin, Enterprise, Admin Controls, and that's in three months. That's the pace we're on. Now project that forward. What does it look like in six months?

What does it look like in a year? I think by the end of 2026, the average knowledge worker will have an AI agent that handles 30 to 40% of their routine tasks. Not because the technology isn't ready. It's ready now, but because adoption takes time.

People need to trust it. Organizations need to roll out. Workflows need to adapt. But the early adopters, the people who start using this stuff now, they're going to have a massive head start.

And I want to acknowledge something though, because I think it's important. This technology is moving fast. It's going to change how work gets done. See, some people are going to lose their jobs because of this, right?

I'm not going to pretend that way. If your job is primarily like compiling information or summarizing documents or managing schedules or routine communications, you need to be paying very close attention right now because an AI agent can now do all those things. And it doesn't need sleep, health insurance, or a parking spot. But here's the thing that I keep coming back to, right?

Every major technology shift creates more jobs than it destroys. It just creates different jobs. The people you learned to build websites in 1998 became the digital agency owners of 2005. The people who figured out social media marketing in 2010 became the influencer economy of 2020.

The people who figure out AI agents in 2026 are going to build the next generation of businesses and careers. And this is already happening, right? So here's what I'd actually do if I was you. If you're not technical, download Cloud Desktop and just check it out and test out some prompts on it.

If you're a little bit more technical and you want more power, look into Cloud Code. It's in the terminal, understands your entire code base. It can write features and all sorts of cool stuff. If you're running a business, start thinking about which of your teams for repetitive tasks could be handled by co-work with connectors and plugins, email triage, and report generation.

And if you're running a big organization, talk to Anthropic about their enterprise plan, right? Or look at what Microsoft just built with Copilot Co-Work, the cloud-based version that sits inside your Microsoft 365 tenant with full enterprise data protection. The on-ramps are everywhere now and there's no excuses left, right? And I've been covering AI on this channel for a long time, and I can tell you the weeks blur together sometimes.

There are always new announcements, another model, another feature, but every once in a while, something drops that you can feel is different, right? It's something that actually changes the trajectory, not for developers, not for just early adopters, but for everyone. A dispatch is one of those moments, not because the technology is revolutionary. Honestly, OpenClaw proved this concept months ago, but because the accessibility just crossed a threshold.

And when get an AI agent goes from a weekend project to a two minute setup, you've crossed from early adopter territory into mainstream territory. Mainstream is where the real money is. Mainstream is where the real impact is. Mainstream is where the world actually changes.

Two downloads, one QR code, the AI agent live. That's a new reality and it happened this week. The people who are going to thrive in this new world aren't the ones who know the most about AI. They're the ones who actually use it, who actually set it up, who actually delegate their first task and feel what it's like to come back to finished work.

Don't be the person who watches this video and thinks, Oh, that's interesting. And then goes back to doing everything manually. Be the person who pauses this video, downloads the app and has an AI agent running before dinner tonight, because the gap between the people who use this stuff and the people who don't is about to become a canyon and you get to choose which side you're on. If you want to check out more and learn more, check out the AI Profitable Boardroom, where we cover this sort of stuff.

We cover practical tutorials on exactly how to use it, and we show you how to automate your business and save time and scale using automations in a very practical way. Thanks for watching. I'll see you in the next one.