Anthropic Reverses Ban on Third-Party AI Agent Use

Anthropic has reversed its April 2026 policy banning third-party AI agents on Claude subscriptions, announcing a new Agent SDK credits subcategory for paid subscribers. These credits can now be allocated specifically for programmatic uses, including external tools like the open source OpenClaw agentic harness. The original ban came after Anthropic cited capacity and service issues caused by subscribers using $20 to $200 monthly plans to consume far more resources than intended through third-party agents. The reinstatement comes with a catch, though details of the new credit allocation system suggest limits on how these credits can be used.