Here it comes—Anthropic has taken action against OpenClaw. Starting today at 12 PM Pacific Time, the era of using Claude subscription quotas to run third-party tools is officially over. OpenClaw is the first framework to be called out, and this will eventually extend to all third-party tools. In other words, if you were previously paying $20/month for a Pro subscription to run your own AI agent system, that path is now blocked. Subscriptions will only cover Claude’s own products (Claude Code, Cowork), while third-party frameworks will be charged separately based on actual usage. Someone previously calculated that the $200 Claude Max usage roughly equates to $5,000 worth of API usage. Anthropic’s reasoning is straightforward: these tools were putting too much strain on the system, and production capacity needs to prioritize core product users. As compensation, they’re offering a one-time usage credit (equal to the monthly subscription fee), redeemable before April 17. They’ve also introduced prepaid usage bundles with discounts of up to 30%. However, compared to subscription quotas, additional usage burns through credits much faster. To put it bluntly, too many people were gaming the system, and Anthropic crunched the numbers to find that agent users’ token consumption was completely disproportionate to their subscription fees. This is a step all AI companies will inevitably take—first attract users with unlimited plans, then switch to usage-based pricing once they’ve gained the upper hand.