Anthropic Restores Third-Party Agent Access, but Imposes Budget Limits

Anthropic recently announced via its official developer account that it is updating its Claude paid subscription plans, reintroducing the ability for subscribers to use third-party AI agents like OpenClaw to access Claude models. The company had previously cut off this feature due to resource pressure, sparking widespread community concern. Under the new policy, all Claude Pro ($20/month) and Max ($100–200/month) subscribers will receive a dedicated "Agent SDK" credit to cover the costs of programmatic calls.

This restoration is not unconditional. The key change: third-party agent calls will no longer be subsidized by the general subscription pool, but instead limited to a fixed, non-rolling monthly credit allowance, with amounts corresponding to subscription fees ($20–200) and billed at standard API rates. Previously, some users consumed far more computing resources than their subscription fees via autonomous agents like OpenClaw, straining the system. Anthropic said the new mechanism aims to balance user experience with resource fairness and ensure service stability.

The move has sparked mixed reactions in the developer community: on one hand, heavy automation users will need to pay for overage; on the other, lightweight users can continue to integrate third-party tools through their subscriptions. Industry observers believe this marks a shift for AI model providers from an "unlimited usage" model to more granular resource management, and similar credit-based designs may become mainstream in the future.