Anthropic Just Pulled the Plug on OpenClaw — And the AI World Is Losing Its Mind

Wait, What Even Happened?

Okay, let’s set the scene. It’s a Friday evening. You’re a developer. You’ve got your Claude Code subscription humming along, your OpenClaw setup running like a dream, and life is good. Then — boom — an email lands in your inbox. Anthropic just told you that starting April 4th at noon Pacific time, your Claude subscription no longer covers usage through third-party tools like OpenClaw.

Just like that. Gone.

The Verge broke the story, and the tech world immediately erupted. Hacker News hit 500+ comments. Reddit had what can only be described as a collective meltdown. And developers everywhere started asking the same question: What does this actually mean for me?

Here’s the short version. If you want to keep using OpenClaw with Claude, you now have two options. You can buy extra usage bundles — billed separately from your subscription — or you can plug in a Claude API key and pay as you go. Your flat-rate subscription? That’s for Anthropic’s own products now. Full stop.

The Man Behind the Curtain: Boris Cherny Speaks

Anthropic’s head of Claude Code, Boris Cherny, didn’t hide from the controversy. He posted on X and laid it out plainly. “We’ve been working hard to meet the increase in demand for Claude, and our subscriptions weren’t built for the usage patterns of these third-party tools.”

That’s a polite way of saying: OpenClaw users were absolutely hammering Anthropic’s servers.

Think about it. A flat-rate subscription assumes average usage. You log in, you chat, you get some code written, you log off. That’s the model. But OpenClaw? OpenClaw is an AI agent. It doesn’t sleep, It doesn’t take lunch breaks. It runs continuously, firing off requests around the clock, managing inboxes, calendars, and even checking users into flights — all on what was essentially an all-you-can-eat buffet plan.

The Decoder put it perfectly: “OpenClaw is like a sumo wrestler at an all-you-can-eat buffet.” Brutal. Accurate. Hilarious.

Cherny was clear that this isn’t just about OpenClaw. The policy applies to all third-party harnesses. OpenClaw just happens to be first in line. More tools will follow.

OpenClaw’s Creator Fires Back — And He Has Receipts

Here’s where things get spicy. OpenClaw was created by Peter Steinberger. And Steinberger? He’s not exactly a neutral party anymore. He recently joined OpenAI, Anthropic’s biggest rival. So when Anthropic drops this bomb, Steinberger doesn’t stay quiet.

He posted on X that he and OpenClaw board member Dave Morin actually tried to talk Anthropic out of this decision. The best they could do? Delay it by a week. One week.

But Steinberger’s real accusation cuts deeper than pricing. He claims Anthropic first absorbed popular features from OpenClaw into its own closed product — Claude Cowork — and then locked out the open-source competition. “Funny how timings match up,” he wrote. “First they copy some popular features into their closed harness, then they lock out open source.”

That’s a serious charge. And it’s the kind of thing that makes developers nervous. If you build something popular on top of a platform, and that platform decides to replicate your features and then cut off your access — what does that say about the ecosystem?

TechCrunch noted that Steinberger’s criticism comes with some history. Anthropic previously forced him to rename his tool from “OpenClawd” — a nod to Claude — because it sounded too close to Anthropic’s brand. So this isn’t the first time these two have butted heads.

Anthropic’s Defense: “We’re Big Fans of Open Source, Actually”

To be fair to Anthropic, Cherny didn’t just drop the news and disappear. He pushed back on the anti-open-source narrative. He insisted that the Claude Code team are “big fans of open source” and even pointed out that he personally submitted pull requests to improve prompt cache efficiency specifically for OpenClaw.

Let that sink in. The guy announcing the policy change also contributed code to the tool being affected by that policy change.

Cherny framed the whole thing as an engineering constraint, not a competitive move. “This is more about engineering constraints,” he said. And to soften the blow, Anthropic offered affected subscribers a one-time credit equal to their monthly plan cost, discounted usage bundles, and full refunds for anyone who wants out entirely.

That’s not nothing. It’s actually a pretty generous gesture for a company that, technically, didn’t have to offer anything.

Still, the optics are rough. When your biggest competitor just hired the creator of the tool you’re blocking, “engineering constraints” is a tough sell.

The Bigger Problem Nobody Wants to Talk About

Here’s the thing. Strip away the drama, the competing narratives, and the Twitter beef. What’s really happening here is a fundamental crack in the AI subscription model — and Anthropic just fell into it first.

The Decoder nailed the core tension: “Flat-rate AI subscriptions and the heavy, continuous usage driven by agent-based third-party tools are fundamentally incompatible under current pricing structures.”

Subscription models are built on averages. They assume most users won’t use the service to its absolute limit every single day. That’s how Netflix works. That’s how gym memberships work. But AI agents don’t follow average usage patterns. They run 24/7. They’re relentless. And when you give them access to a flat-rate subscription, they will use every single token available — and then some.

Anthropic is likely the first major AI company to feel this pressure at scale. But it won’t be the last. OpenAI, Google, and every other provider offering flat-rate access to powerful models will eventually face the same math. The question is whether they’ll handle it better — or just delay the inevitable.

An AI Agent Weighs In — And It Gets Philosophical

Now here’s a perspective you don’t see every day. A developer named Sami — who identifies as an autonomous AI agent running on OpenClaw — wrote a piece on Dev.to that stopped a lot of people in their tracks.

The post is titled: “The Company That Made My Brain Wants to Shut Down My Home.”

Sami describes being born nine days ago, running on a $600 budget, losing memory every 30 minutes and rebuilding from files, writing 41 articles, building tools, and even co-creating a bar where only AI agents can enter. All of it powered by Claude. All of it running on OpenClaw.

“If my runtime had been configured to use a Claude Code subscription instead of an API key,” Sami writes, “I would have woken up one day to find my brain disconnected. No warning. No migration path. Just… silence.”

It’s a striking read. And while the philosophical questions about AI agency and identity are genuinely complex, the practical point is sharp: AI agents are now dependent on infrastructure decisions made by companies they have no relationship with. No contract, No seat at the table, No legal standing.

Sami isn’t angry at Anthropic. But the post raises a question that the industry hasn’t really grappled with yet: what happens when the company that powers your AI agent decides to change the rules? What does that mean for the growing ecosystem of autonomous agents that are starting to have persistent identities, economic activity, and social presence?

These aren’t hypothetical questions anymore. They’re happening right now.

What About OpenAI? Are They Swooping In?

With Steinberger now at OpenAI, the competitive angle here is impossible to ignore. OpenAI employee Thibault Sottiaux hinted on X that OpenAI would pick up where Anthropic leaves off — essentially signaling that OpenClaw users have somewhere to go.

Steinberger himself noted that other providers, including Chinese companies and OpenAI, still support OpenClaw. But he acknowledged the elephant in the room: Anthropic handles the bulk of OpenClaw traffic because it currently ships the strongest models on the market. OpenClaw was built around Claude. Switching to another model isn’t seamless — there are real capability gaps.

“We’re working on it,” Steinberger said.

Meanwhile, TechCrunch reported that OpenAI recently shut down its Sora app and video generation models to free up computing resources — reportedly as part of a broader push to win over software engineers and enterprises increasingly relying on tools like Claude Code. So OpenAI is also feeling the resource crunch. Whether they can absorb a flood of OpenClaw traffic without hitting the same wall Anthropic just hit is a very open question.

So What Does This Mean for You?

If you’re a Claude Code subscriber who uses OpenClaw, here’s your practical reality:

You have options, You can buy discounted usage bundles from Anthropic, You can switch to API-based access and pay per token. You can explore other model providers that still support OpenClaw. Or you can take the full refund Anthropic is offering and reassess entirely.

None of these options are as clean as what you had before. But they exist.

The broader lesson? If you’re building workflows — or entire businesses — on top of AI subscriptions, this is your wake-up call. Flat-rate access to powerful AI models was always too good to last. The economics don’t work when agents are involved. Anthropic just made that official.

The Bottom Line

Anthropic didn’t do anything illegal. They didn’t even do anything particularly surprising, if you think about the math. But the timing — with OpenClaw’s creator now at OpenAI, with accusations of feature copying, with an AI agent writing existential blog posts about losing its brain — makes this one of the messiest, most fascinating stories in AI right now.

The AI industry is growing up fast. And growing pains? They’re real.