As of April 4, 2026, Anthropic officially blocked third-party tools like OpenClaw and OpenCode from drawing on Claude Pro/Max subscription limits. If your tool isn’t Claude Code or Claude.ai, your subscription no longer covers it — you’ll need to pay API-rate “extra usage” on top of what you already pay. Anthropic is offering one-time credits ($20–$200 depending on your plan) claimable by April 17, plus bundle discounts on extra usage. Full refunds are also available for subscribers who don’t accept the new terms.
This is the culmination of months of escalation: OAuth blocks in January, a ToS rewrite in February, account ban waves, the massive Claude Code source leak in March, a DMCA blitz targeting 8,100 GitHub repos, and now this. OpenClaw’s creator Peter Steinberger accused Anthropic of copying popular open-source features into their own tools and then locking out the competition.
The core tension: power users running autonomous agents through third-party harnesses were burning through $1,000–$5,000 in API-equivalent tokens per day on a $200/month subscription. Anthropic’s view is that this was unsustainable. The developer community’s view is that it’s a bait-and-switch — “unlimited” turned out to mean “unlimited, but only through our tools.”
The article frames this as the final brick in a walled garden: leak the code, DMCA the mirrors, ban the rebuilds, block the workarounds, cut the subscriptions. The piece concludes that Anthropic is operating in reactive mode, making policy one crisis at a time.
[AIL4]