As of today, April 4 at 3 PM ET, Claude subscribers can no longer use their subscriptions to connect Claude models to third-party agentic tools like OpenClaw.
What’s Changing
Up until now, users with a Claude Pro or Team subscription could plug their credentials into tools like OpenClaw — a popular platform that orchestrates AI agents across different models. It worked, but it was never officially supported.
Boris Cherny, Head of Claude Code at Anthropic, put it plainly: subscriptions were never built for the usage patterns these third-party tools create. The strain on Anthropic’s systems was disproportionate.
Why Now?
Anthropic says they want to manage capacity ‘thoughtfully’ and are prioritizing their own products and API. Translation: OpenClaw users were burning through compute that Anthropic wasn’t getting adequately compensated for.
Your Options Going Forward
If you still want to use Claude through third-party tools, you’ve got two paths. First, you can buy ‘extra usage bundles’ tied to your Claude login, which come with up to 30% off. Second, you can grab a separate API key through Anthropic’s developer platform — that’s straightforward pay-per-use.
As a consolation, existing subscribers get a one-time credit equal to their monthly plan price, redeemable until April 17.
My Take
This was inevitable. When a $20/month subscription suddenly powers agentic workloads that consume multiples of that in compute, the math doesn’t work for Anthropic. At the same time, it shows how fast the ecosystem around LLMs is evolving — and how providers are struggling to keep up with usage patterns they never anticipated.
For developers and power users, the workflow changes: if you’ve been conveniently using your subscription everywhere, you now need to budget for API usage. That’s fair, but it stings.
Sources: