Anthropic has blocked OpenClaw from using Claude Code subscriptions, and the reason seems to be simple math. They don't have enough compute. OpenClaw is an autonomous tool that hits Claude's API hard, and with Claude Code usage blowing past Anthropic's growth projections, something had to give.
The infrastructure situation is complicated. According to Hacker News discussion, Anthropic apparently bet on smaller neo-cloud providers instead of locking in capacity with the big hyperscalers like AWS or Google Cloud. Some of those partners haven't delivered on time. Meanwhile, Claude Code took off faster than anyone predicted during their last capacity planning cycle. The result has been brownouts and errors, leaving users frustrated.
This is forcing Anthropic to pick winners. Direct Claude Code users get priority. Third-party automation tools like OpenClaw get cut off. Money wasn't the primary issue. OpenClaw users were paying for subscriptions. But an autonomous agent making thousands of calls is fundamentally different from a human developer using Claude Code normally. When you're capacity constrained, you protect your core product.
Some users are already looking elsewhere. Chinese models like GLM and Minimax keep coming up in discussions as alternatives with better throughput and lower costs. For Anthropic, the calculus is brutal: keep the power users happy and risk alienating the ecosystem, or spread capacity thin and watch the core product degrade.