Anthropic Just Killed OpenClaw

0:00So, Anthropic just completely killed Open Claude just a few hours ago, and we're going to talk about in this video what we're going to do about it as builders. If you're running Open Claude and you've got agents running in the background, cron jobs firing off overnight, any of this directly affects you. Starting tomorrow, 2026, yes, tomorrow, the way third-party harnesses like Open Claude that access Claude through the OAuth is completely different. Not saying you can't do it entirely, but the way you are now charged changes dramatically. So, different pricing, different billing, a completely different game. So, I'm going to break down exactly what changed and why Anthropic did what it did and what the community is saying about it. What I personally changed in my setup because

0:46it is going to be helpful for how we operate going forward, and maybe can shed some light on how you're going to configure your whole workflow. I have 17 agents running 24/7 on Claude in Open Claude, and this affects me directly. So, let's start with what actually changed because if you haven't been following this closely, it might feel like it came out of nowhere. We all expected this to happen eventually. There's no way that Claude or Anthropic can offer two to $3,000 worth of tokens on a $200 month plan. That doesn't work, right? So, this has been building since January. And here's how it worked before if you were using your subscription through Open Claude. Let's say you had a Claude Max subscription. That's a flat monthly rate, right? You could use your OAuth to authenticate through that

1:32subscription within Open Claude. So, your agents, your cron jobs, your background tasks, they were all writing on your subscription on one flat fee, and it felt like a great deal. You can build a lot of stuff for cheap. So, what Anthropic did starting tomorrow is they split that. Your Claude Max subscription still works for Claude code directly. For the actual Claude code CLI tool, It works, but for third-party harnesses, apps like Open Claude that use the OAuth to connect, they now have to go through a separate metered billing. And don't forget to claim your free credit. You get matched whatever your subscription was as a free credit. There's that at least. So, meaning basically now, you pay per token, per API call separate

2:18from your subscription. Real quick, before we keep going, if you're watching this and you want to actually build with some of these tools, not just watch videos about them, then you're going to want to check out our community down below, Shipping School. We have a full Claude Code course, a full Open Claude course, and four live bootcamps every single week where we actually help you get set up from scratch. Like actually set this thing up, not just watch a tutorial and figure it out by yourself. And we also provide one-on-one coaching, so you could book a call with me, we could share screens, and I can help you get Claude Code or Open Claude running on your machine. That's it. No fluff. I built this community because watching YouTube's only gets you so far. We launched it just 3 days ago and we have over 55 members. You need people around

3:05you who are actually building, people who hold you accountable, and coaches who can help you when you get stuck. I'll put the link in the description down below. Get in now before the price goes up. So, that's the change in plain English that I could break it down in simple terms. So, one lane for Claude Code that you could still use within VS Code, WinSurf, and Cursor. That's doesn't change, and that's going to be an amazing experience if you're building with it tomorrow. And now, there's a separate lane for users who use Claude within Open Claude. And that second lane has now metered pricing. Why did they do this? Because the economics were completely broken. If you think about this, imagine you buy an all-you-can-eat buffet pass and you figure out a way to run the catering company off that pass, like loading up industrial containers,

3:50running off full-service dinners for other people every single night and the buffet is like wait, that's not what I meant. You're taking advantage of the system. That's essentially what was happening. A human sitting down to use cloud code, they're coding for a few hours, they hit some token limit, they stop. Normal usage. A third-party harness running autonomous agents, there's no human there hitting a stop button. The agents run consistently. They run on a loop, right? They process, they don't have natural speed limits that humans do at typing at keywords and ideating and whatnot. VentureBeat actually put it this way. Third-party harnesses remove those speed limits and that's exactly right. When I had 17 cron agents running, some of them firing every 30 minutes all night, that's a completely different consumption pattern

4:35than one developer using cloud code to just ship a feature. The Register reported on this back in March and just on the 31st, like 4 days ago, Anthropic was actually admitting that cloud code quotas were running out way too fast and everyone had been complaining about this. They were reducing quotas during peak hours because the load was way higher than they expected and a big chunk of that load was coming from exactly this. Agents running through third-party harnesses without any of the natural friction that comes with human usage. So, Anthropic Anthropic basically looked at their infrastructure costs and said these are two fundamentally different products. We need to price them as two fundamentally different products as so and that's what tomorrow is leading up to. Want to talk about the community reaction because it's been a

5:21heated debate and unfortunately, I just got my Anthropic tattoo yesterday, so there's no going back. Probably not a good decision. So, on the open claw subreddit and on Hacker News, people are generally frustrated and I get it. The frustration isn't really about the pricing change itself, it's about the communication. People feel misled because Anthropic previously said, they said that cloud code OAuth was fine for use with open claw. They said it was okay and then they reversed that. The Hacker News framing I saw that stuck with me was basically this: Claude code is a lock-in play, like Apple's App Store. You build in the ecosystem, you get comfortable, you depend on it, and then the rules change and you're stuck paying whatever they decide to charge. That's a huge concern, and I'm not dismissing that at all. But here's where

6:06I land on it after sitting with it and just kind of marinating. Anthropic is a company, right? They have costs. Those costs got way higher than planned because the use case changed. Nobody anticipated that people would be running full autonomous agent fleets through an O off subscription designed for a single developer tool. That wasn't the original contract, even if it wasn't explicitly prohibited. It's very frustrating, I know, when the tool when the rules change, right? And you're playing the game. It's not fair, right? Oh my goodness. But it from a business point of view, it doesn't make sense. It's unsustainable. You know, and I'm being honest and straight-up here with you. Uh the original deal was kind of too good to be true to begin with. Flat rate unlimited background for no overage for $200? That's insane. That was never sustainable. And the part that's

6:51legitimately fair to be upset about is just the communication. Just to be clear from the start. Don't say one thing and then change it later. That erodes the trust, and Anthropic needs to do better there. But the pricing change itself, I mean, it does make sense. Okay, for my personal situation, here's what I'm doing. Like I have 17 agents all doing different things on cron jobs and with an open clock. And some of them do run every 30 minutes around the clock. Content pipeline agents, the memory sync with the heartbeat, research runners, social media posting, all of it, right? And most of that was through Claude. So, when this change started getting more telegraphed in January, I started shifting things around and trying new models out. And by the time this day came, we all knew it would come, I was mostly already in a different configuration. And I'm happy where I'm

7:36at now. So, what I did is I switched my main chat, the actual back and forth where I'm talking to Atlas, the coordination nation, the one that executes everything, to Sonnet 4.6, and I'm paying the usage charges, which is fine. I don't mind. And it's still Claude, and it's great for the conversation. I'm paying for that cuz I notice a huge difference between GPT 5.4 in the main chat versus Anthropic models. I will pay extra to have good communication and more personality. But, here's the thing. I used to run Opus 4.6. I can't do that right now. And I'm not going to want to be burning Opus credits for that reason. And I'll talk about how I'm going to use Opus kind of within Open Claude later. For the Chron agents specifically, all 17 of them, I moved them all to GPT 5.4 on the OpenAI

8:24OAuth took. Separate metered costs, and honestly, a lot of them were already on OpenAI because I'd been doing the model routing work before this really even happened. I added a $100 usage credit buffer, just a buffer sitting there so nothing goes dark unexpectedly. In case I usage spikes. It is more expensive than before, a little bit, yes, but it is predictable now. And predictable beats cheap but fragile every single time. I can't be, you know, cutting costs right here. Like this is a machine that needs to run. If you're struggling to keep up with content, well, I'm about to save you about 40 days worth of work. I built something [music] called Content Machine. It's 10 AI agents that run on the Open Claude orchestration, and they handle everything. Scripts, thumbnails, X posts, blogs, outreach, clips,

9:10newsletters, all of it. So, I went from 1,000 subscribers to 4,000 subscribers on YouTube in 7 days using this exact system. Every single morning, I wake up, >> [music] >> and the content's already done. I spend maybe 15, 20 minutes reviewing and approving them, and I move on with my day. It works for any niche, fitness, [music] finance, real estate, marketing, whatever you are building, and it is 100% completely customizable to your use case. So, you get the mission control dashboard, all of the cron jobs, everything I've built over the last 40 days helping me gain more and more people to subscribe and join the community. So, you plug in your own thing and it molds it to you, it learns how you talk and it writes so it doesn't

9:55sound like AI slop. $97 one time, it's not a subscription. I'll put the link down below and you'll thank me later. Here's the question I actually explored and I want to give you you know, a straight-up answer. Um like my question was could you just use Claude Code inside Cursor to control OpenClaw instead? It's a good question, right? Like skip OpenClaw's native authentication and just pipe everything through Cursor's Claude Code integration. I went down this road and the short answer is yes, for some things, but not as a replacement. So, for example, like surgical tasks, it totally works. You open Cursor, you're running Claude Code inside of Cursor, you can use VS Code or WinSCP, whatever you want. Right? You just use Claude Code in the terminal. But you you can make a specific target change to your

10:40workspace in OpenClaw and I love that, but you can't replace OpenClaw with it. Here's why. OpenClaw has heartbeats, it has background agents that run on a schedule without you touching anything. It has Telegram integration, so you get notified when things happen. It has memory that persists across the session. It has cron jobs that fire while you're sleeping. Claude Code inside of Cursor is reactive. You open it, you ask it something and it does the thing and then you close it. That's it. There's no run this agent every 30 minutes and message me on Telegram if something's wrong. That's not what Claude Code is. So, the question is can Claude Code replace OpenClaw? It's kind of like asking a really good power drill to replace a construction crew. The drill is an incredible tool and Claude Code has changed my life and you can use it constantly, but it doesn't show up at

11:267:00 in the morning, it doesn't run the whole project and it doesn't text you when there's a problem. There's really the core takeaway here and I want to make sure this lands with you. Open Claw is a 24/7 autonomous system. It runs when you're not there. It runs when you're asleep. It has agents that monitor things, create things, and post things, and sync things, all without you actively sitting there prompting it. Cloud Code is just a tool, an insanely powerful tool. You use it to build, to fix, to ship things, but it's reactive. You have to activate it. It responds and then it is done. These are not competing products. They're not even in the same category. They do different things for different purposes. And Anthropic separating the billing is actually just acknowledging that reality, straight up. And a 24/7 autonomous agent fleet across

12:14more is going to cost more to run than a developer's coding assistant. That's just not an opinion. That's just math being math, right? So, if you're building on Open Claw, here's what I'd tell you. Don't panic about this. Don't audit your agents. Know which ones are actually driving value. Don't run 17 agents if 12 of them are dead weight. Pay for what you actually use. Optimize your workflow. And if budget is a real concern, look at routing your background crons to something like OpenAI, even using the API or your auth token, or using one of new Google's models, or checking out open weight models on open router. Experiment. GPT-5.4 handles most of it pretty well and the cost is predictable. The tools are still there. Their capability is still there. The game just got a little bit more honest

12:59about what things actually cost. So, if you're the kind of person who's building this stuff, who's like running agent setups, who's thinking seriously about autonomous AI systems as part of how you work, I want to be building alongside you, which is why we have a community of 197 people that we launched just 20 days ago. It's called Shipping School, where we figure this out stuff together. Real people, real building real real things. And it's not just theory, it's not hype, you know, we've got eight live calls a week. We're working through exactly these kind of questions, what models to route to, where and how, and how to build agent pipelines that actually hold up, how to optimize your workflow, how to start a business making money with agents as a service. And I'll put the link down below. Come build with us, check it out. We do have courses in there, but we're focusing on actually

13:45group and community and human-to-human interaction and eight live calls a week. So, if you haven't subscribed already, please subscribe to the channel. I release about four to six videos like this every single day covering open claw use cases, AI news, and anything claw code related. See you guys in the next one. Have a blessed day.