Anthropic Bans OpenClaw From Claude Subscriptions, Forces Users to Pay Extra

AI interface being disconnected symbolizing Anthropic cutting off OpenClaw access

If you have been using OpenClaw with your Claude subscription, your free ride ends today. Anthropic announced Friday evening that starting April 4th at 3PM ET, Claude subscriptions will no longer cover usage from third-party tools like OpenClaw. Users who want to keep using OpenClaw with Claude will need to pay extra through a separate pay-as-you-go billing option.

What Is Changing

Anthropic's Claude Code exec Boris Cherny laid out the new rules: Claude subscriptions were never designed for the usage patterns that third-party tools like OpenClaw create. The company is now "prioritizing customers using our products and API" — a diplomatic way of saying OpenClaw users were consuming too much capacity at subscription rates.

To soften the blow, Anthropic is offering subscribers a one-time credit equal to their monthly plan cost and discounted usage bundles. Users who want a full refund can request one via a link in tomorrow's email.

The OpenClaw Backstory

OpenClaw exploded in popularity earlier this year as an AI agent that could efficiently manage inboxes, calendars, and even check into flights. The tool quickly became one of the most popular ways to interact with Claude — and apparently put a significant strain on Anthropic's infrastructure.

What makes this move particularly interesting is that OpenClaw creator Peter Steinberger is now employed by OpenAI. According to Steinberger, he and OpenClaw board member Dave Morin "tried to talk sense into Anthropic, best we managed was delaying this for a week."

The Open Source Backlash

The reaction from the developer community has been swift and critical. Steinberger noted on X that Anthropic "first copied some popular features into their closed harness, then locked out open source." The implication is clear: Anthropic built competing features into Claude Cowork (their own tool), then cut off the third-party alternative.

Not everyone agrees with the criticism. Developer Dex Horthy argued the move is "well within TOS" and that "there was never any promise to give people cheap inference." He pointed out that before the Claude Code max plan, everyone was paying per token anyway.

A Pattern in AI

This follows a familiar playbook in tech: platform allows third-party innovation, third-party gets popular, platform builds competing feature, platform restricts third-party access. Twitter did it with its API. Apple does it with App Store rules. Now AI companies are doing it with inference capacity.

The difference here is that AI inference is genuinely expensive. Anthropic giving away cheap Claude access through subscriptions while OpenClaw users consumed disproportionate compute was arguably unsustainable. But the timing — building a competitor first, then pulling the plug — makes it hard to frame as purely a capacity issue.

The Bottom Line

Anthropic has every right to manage its infrastructure and pricing. But the way this played out — copying features, hiring away the competition, then cutting access — feels less like "thoughtful capacity management" and more like a classic platform squeeze. OpenClaw users now face a choice: pay more, switch to Claude Cowork, or find another AI provider entirely.