At noon Pacific today, if you were using Claude through OpenClaw or any third-party AI agent framework, your subscription stopped working. Not broken — intentionally switched off.

Anthropic announced the change earlier this week. Starting April 4, Claude Pro and Max subscribers can no longer use their included usage limits with third-party harnesses like OpenClaw. They'll need to pay separately, via a new pay-as-you-go billing tier billed outside their existing subscription.

OpenClaw is one of the most popular open-source AI agent frameworks. Developers use it to chain Claude into automated pipelines, build persistent personal AI assistants, and run long-horizon coding workflows that the Claude.ai interface wasn't built for. The tool's creator, Peter Steinberger, and board member Dave Morin personally lobbied Anthropic to reverse the decision. They succeeded in delaying enforcement by exactly one week. That was the full extent of their leverage.

Boris Cherny, Head of Claude Code at Anthropic, wrote the announcement. "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." Third-party harnesses were placing an "outsized strain" on its infrastructure, the post said, and capacity must be prioritized for customers using its core products.

The consolation offer: a one-time credit equal to your monthly subscription cost, redeemable by April 17. Plus a 30% discount on pre-purchased usage bundles — if you want to keep using Claude at all via API-backed tools. Enforcement starts with OpenClaw and will expand to all third-party harnesses in the coming weeks.

This isn't a capacity story. Anthropic has the infrastructure to serve these requests. What they don't have is a pricing model that makes it profitable. When a developer runs 50 automated tasks per day through OpenClaw on a $20/month Claude Pro subscription, the unit economics collapse at scale. That's real. But framing a pricing correction as an infrastructure protection measure is a choice — and it's the kind of choice that reveals what a company actually thinks of its developer community.

The power users getting cut off today aren't churners. They're the people building workflows on top of Claude, writing tutorials, evangelizing the product in Slack channels and Discord servers. They chose Claude over GPT-5.4 for their stacks. And they'll remember how this felt.

My Opinion

I'll be blunt: this is a bad look for Anthropic, and they know it.

The developer community is exactly the group Anthropic needs on its side right now. OpenAI is shipping GPT-5.4 with native computer use. Google is pushing Gemma 4 for agentic workloads. The AI wars are increasingly being won or lost in the tools developers actually reach for. Cutting off subsidized access — with a week's notice — over what amounts to a pricing miscalculation is the kind of move that turns evangelists into skeptics overnight.

Yes, I understand the math. Heavy automation users were almost certainly burning through API capacity on a flat-rate ticket. That needed to change. But there's a respectful way to do that — a dedicated developer tier, a metered add-on, a transparent conversation about costs — and there's a cliff-edge policy announcement. Anthropic picked the cliff.

Here's what I think actually happens next: OpenClaw developers migrate toward OpenAI's API or Gemini, just as a point of principle. Anthropic gets its capacity back but loses a cohort of builders who would've been its most vocal advocates. The one-time credit softens the short-term anger. The long-term trust damage is harder to undo. Watch whether Anthropic introduces a proper developer tier within the next 60 days — that would signal they heard the feedback. If they don't, this policy becomes a permanent marker of where their priorities actually lie.

Author: Yahor Kamarou (Mark) / www.humai.blog / 04 Apr 2026