Anthropic sagt, dass Claude-Abonnements OpenClaw nicht mehr unterstützen werden, weil es eine „übergroße Belastung“ für die Systeme darstellt

    https://www.businessinsider.com/anthropic-cuts-off-openclaw-support-claude-subscriptions-2026-4

    Share.

    16 Kommentare

    1. Is it actually this case or is it just that OpenAI bought OpenClaw and Anthropic want nothing to do with supporting OpenAI?

      I’m quite sceptical about it all because I’ve yet to receive any of the notices or emails that are supposedly going out about the usage restrictions.

    2. No-Contract9167 on

      What makes this interesting is that it exposes the pricing-model mismatch more than anything else. Flat subscriptions work fine until a small slice of users turns the product into near-continuous agent infrastructure. At that point the company either raises prices, introduces usage tiers, or starts carving out high-intensity workflows.

      The broader takeaway is that consumer chat pricing and agentic automation pricing probably can’t stay bundled forever. They behave like two different products with two very different cost curves.

    3. Yes, and this exposes how fragile their cost models are.

      The „app“ level APIs have much higher token limits compared to the per request APIs they sell on the market.

      I’m paying $10(?) to Gemini, the amount of queries I did would easily cost $100 per month, or more. (I know, because we pay for Claude, which is per request)

      So… when people take those App tokens and use elsewhere, they really burn money. And money they don’t ever expect to see a return. At least in their App they sell ads and brand recognition.

      (Gemini + AntiGravity has a similar thing. People stole the IDE tokens to use with Claw just to have their accounts banned)

    4. trilobyte-dev on

      OpenClaw is really wasteful with tokens, more so than just about any other use case I’ve seen.

    5. zoupishness7 on

      I didn’t use OpenClaw, but I think this applies to all agent harnesses. So, I just spent $400, to develop my own harness, that I now can’t afford to use on their platform. Good thing I built in Gemini support.

    6. IntelArtiGen on

      I’m not surprised, and I wouldn’t be surprised if it’s not true only for Claude but also for everything we can find online. Most websites did well when most users were humans. When my grandma will be able to scrap 10 websites with a single prompt I’m not sure the internet will support that very well. We will see more paywalls, more „are you a robot?“, etc.

    7. Selling „unlimited“ plans that actually have usage limits should be illegal.

    8. Arent they just enforcing their own TOS? Use cases like openclaw is what your api key is for.

      the weird thing with anthropic is only that they have a prepaid api plan and have to be approved for monthly billing.

    9. Usefulness of open-weights models and costs of on-prem inference are converging. If they’re too greedy, they may find themselves losing the mid to low tier token sales to on-prem inference. Many agentic workflows need access to sensitive information, on-prem is going to be attractive for some users.

    10. lol-its-funny on

      Here is a concept nobody has heard about. Remember you heard it on Reddit first.

      Rate limiting 🤯

    Leave A Reply