Anthropic kills free Claude rides for third-party tools

Anthropic’s Claude just got a little less generous—and a lot more businesslike. As of 3 PM ET on April 4, the company shut off free access to its AI through third-party wrappers like OpenClaw, a free, open-source assistant for automating emails, calendars, and inbox management. The change, announced by Boris Cherny (Anthropic’s head of Claude Code) on X, means users must now either purchase an extra usage bundle or wrangle a Claude API key—neither of which are free.

This isn’t a mass-market shakeup. Most Claude users, after all, just ask it to summarize documents or draft polite emails. But the move targets a small but influential cohort: developers and power users who’d been leveraging Claude’s brains through tools like OpenClaw (built by the team behind Moltbook) to automate workflows without paying Anthropic directly. The company’s framing is clinical—‘subscriptions no longer cover third-party use’—but the subtext is clearer: free rides for intermediaries are over.

The timing is telling. Anthropic’s recent funding rounds and enterprise push suggest a pivot toward monetizing every access point, not just the front door. OpenClaw’s users, meanwhile, are now staring at a paywall where there once was a loophole.

For the open-source community, this is a familiar story: a promising integration gets popular, then gets priced out. OpenClaw’s GitHub issues page already shows the first waves of frustration, with users debating workarounds or migrating to alternatives like LocalAI or self-hosted Llama instances. The shift also highlights a reality gap: Anthropic’s benchmark claims about Claude’s capabilities often focus on raw performance, but real-world deployment—especially through third-party tools—has always been a gray area.

The bigger question is whether this is about cost control or competitive positioning. Anthropic’s API pricing isn’t cheap, and third-party tools like OpenClaw were effectively subsidized by Claude’s subscription model. Cutting them off could reduce ‘unintended’ usage—or it could be a calculated nudge toward Anthropic’s own enterprise solutions. Either way, the message is clear: if you want Claude’s smarts, you’ll pay Anthropic directly.

Developers, of course, will adapt. Some will switch to open-source models; others will grumble but pony up for API keys. But the move underscores a broader trend: as AI providers mature, the era of ‘permissive by default’ is ending. The free tier was never infinite—it just took a while for the bill to arrive.

In other words, this isn’t about Claude getting smarter—it’s about Anthropic getting paid. The real innovation here is the invoicing.