If you woke up to find your OpenClaw assistant unresponsive, or if you got an email from Anthropic that made your stomach drop, you are not alone. Anthropic has officially cut off OAuth access for Claude Max subscribers using third-party tools, and that includes OpenClaw.
Here is what happened, why it matters, and exactly what to do next.
What Changed
Anthropic updated its terms and Claude Code documentation to explicitly prohibit the use of OAuth tokens from Free, Pro, and Max subscriptions in any third-party product. The language is unambiguous: “Using OAuth tokens obtained through Claude Free, Pro, or Max accounts in any other product, tool, or service is not permitted and constitutes a violation of the Consumer Terms of Service.”
This has been rolling out in stages. Enforcement began as early as January 2026, with a formal documentation update in February and a mass email to affected subscribers in late March. As of now, the change is fully in effect. If you were connecting your Claude Max subscription to OpenClaw through OAuth, that connection no longer works.
To be clear, Anthropic is not blocking Claude models from being used in OpenClaw. Opus, Sonnet, and Haiku all still work. What changed is how you authenticate. The “log in with your Claude subscription” path is closed. Going forward, you need your own API key.
Why This Matters
Many OpenClaw users chose the Claude Max plan specifically because it felt like an all-you-can-use pass. At $200 a month, piping that access into a personal AI assistant made economic sense. That math no longer works under the new terms.
Anthropic’s reasoning comes down to economics. Subscription plans were priced for interactive use on claude.ai and Claude Code, not for powering autonomous agents that can burn through tokens at a much higher rate. Third-party agent frameworks like OpenClaw, and even Anthropic’s own Agent SDK, are now explicitly outside the scope of consumer subscriptions.
The developer community has not been quiet about this. Reactions online have ranged from frustrated to genuinely angry, with many pointing out that the rules changed after people had already built workflows around the old arrangement. That frustration is valid. But the policy is final, so the practical move is to adapt.
What to Do Right Now
Anthropic is offering a $200 free API credit to affected users. This is real money and it will keep your OpenClaw assistant running while you get set up with a proper API key. Here is the step-by-step:
Step 1: Claim your free API credit.
Go to claude.ai/settings/usage and look for the option to claim the $200 credit. This is available to Max subscribers affected by the change.
Step 2: Switch your model in OpenClaw.
Open your OpenClaw chat and type `/model`. Click on Anthropic, then select Sonnet instead of Opus. Sonnet is significantly cheaper per token, which means your $200 credit will stretch much further. On Opus, that credit might last a few weeks of heavy use. On Sonnet, you could get months out of it, and for most tasks, the quality difference is marginal.
Step 3: Set up your own Anthropic API key.
While the credit buys you time, it will eventually run out. Go to console.anthropic.com, create an account if you do not already have one, and generate an API key. Add a payment method so your usage continues uninterrupted once the credit is spent. Then configure that key in your OpenClaw settings.
The Bigger Picture
This is not the end of using Claude with OpenClaw. It is a billing change, not a capability change. Your assistant still works. Your skills, your memory, your workflows are all intact. The only difference is where the bill comes from.
For most users running Sonnet, actual API costs will likely land somewhere between $20 and $60 a month depending on usage, well below what you were paying for Max. So while the transition is inconvenient, the economics on the other side might actually be better.
Take the ten minutes to claim the credit, switch to Sonnet, and set up your API key. Your OpenClaw assistant will be back to full speed before lunch.
