Anthropic is reinstating access for third-party AI agents on its Claude subscriptions through a new metered credit system, effective June 15. The change follows an abrupt April 4 block that cut off tools like OpenClaw from consumer Pro and Max plans, citing unsustainable usage patterns. Paid subscribers will now receive dedicated monthly credits for programmatic calls via the Claude Agent SDK and related tools, separate from interactive limits.
Under the tiered structure, Pro users get $20 in credits, Max 5x plans receive $100, and Max 20x offer $200, with excess usage shifting to pay-as-you-go API rates. Anthropic engineer Lydia Hallie emphasized that subscription prices remain unchanged, positioning the credits as an addition for automation while preserving high-capacity interactive access. Credits do not roll over and must be claimed via account email before the deadline.
The April restriction stemmed from third-party tools bypassing prompt caching efficiencies, creating a compute arbitrage where flat-rate plans powered high-volume agent workflows. Developers expressed frustration over the initial cutoff and new caps, with some Reddit posts calling it a potential 50-fold cost hike for heavy users. Anthropic maintains the system suits individual experimentation, urging production-scale teams to adopt direct API keys.
This adjustment coincides with infrastructure gains, including a May 6 deal granting Anthropic full access to SpaceX’s Colossus 1 data center over 300 megawatts and 220,000 NVIDIA GPUs to double Claude Code rate limits and boost overall capacity. The moves signal Anthropic’s effort to balance developer needs with sustainable scaling as AI demand surges.
Read Article: Apple Developing AI Agent Framework for the App Store: Report
