Anthropic Squeezes Claude Code Power Users
Anthropic sent notices telling Claude Code subscribers that third-party harnesses like OpenClaw are no longer allowed under their subscription. The complaints in the thread are pointed: users are paying for token consumption, and Anthropic is telling them they are consuming too many tokens, which is the thing they paid for. One person compared it to a gym banning members who actually work out.
This connects to a broader pattern of AI platform providers tightening terms as usage scales. Google's Antigravity TOS can get your entire Google account suspended for third-party usage. The story generating real alarm is that the Google account is 'way too precious a thing to lose based on some AI tool.' These are not just pricing disputes. They are existential risks for anyone who has built a workflow or product on top of these platforms.
The thread predicts the next wave: 'agent-agnostic' tools that can swap between Claude Code, other CLIs, and APIs without being locked to any one provider. That is the response the market is generating.
So what?
If you are building a product or internal workflow on top of Claude Code or any single AI provider's subscription tier, you are carrying platform risk that does not show up in your cost model. The smarter move is to architect for provider switching now, before a TOS change forces you to do it under time pressure.