Anthropic to meter Claude agent usage across subscription tiers, alters pricing

Anthropic said on May 14 that it will start metering programmatic Claude agent usage across paid subscription tiers, carving automated workloads out from the broader allowances that had covered chat and coding use. The change, described in reports by Axios and InfoWorld and reflected in Anthropic’s pricing materials, takes effect on June 15 and applies to tools including the Claude Agent SDK, GitHub Actions and some third-party agent frameworks. Anthropic’s paid Claude plans will still include monthly credits for that usage, but once those credits are exhausted, customers will move onto API-style billing. The move changes how developers and power users pay for running Claude in autonomous or semi-autonomous workflows rather than in direct human chats. ### Which Claude activity is being put on a meter? InfoWorld reported that Anthropic will separate “programmatic” Claude usage from standard chat subscription limits starting June 15. That category includes usage through the Claude Agent SDK, GitHub Actions and third-party frameworks such as OpenClaw, according to the report’s description of Anthropic’s blog post. (infoworld.com) Anthropic has been pushing Claude further into autonomous coding and agent workflows for months. The company’s September 2025 product update said the Claude Agent SDK gives teams access to the same core tools, context management systems and permissions frameworks that power Claude Code. ### How much usage comes with each paid plan? InfoWorld reported that Anthropic will assign a monthly programmatic credit pool tied to each subscription tier. (infoworld.com) Under that structure, Pro users receive $20 in credits, Max 5x users receive $100 and Max 20x users receive $200, the report said. Anthropic’s pricing page currently lists Pro at $20 a month when billed monthly and Max plans starting at $100 a month, with a higher Max tier listed at $125 a month if billed monthly. (anthropic.com) The pricing page also says prices and plans are subject to change. ### What changed from Anthropic’s earlier restrictions? InfoWorld reported that Anthropic had said in April, in a post on X, that Claude subscriptions would no longer cover usage on third-party tools such as OpenClaw, citing compute capacity constraints. (infoworld.com) Axios reported on May 14 that Anthropic is now restoring support for outside agent tools on paid Claude plans while tightening other usage limits at the same time. (claude.com) Before that April restriction, InfoWorld said, programmatic workloads and interactive Claude usage drew from the same subscription pool. That allowed some developers to run autonomous agents, scripts and continuous integration workloads under flat subscription pricing rather than through direct API billing. ### Why are developers focused on this change? (infoworld.com) Axios said the dispute shows pressure on “all-you-can-eat” AI subscriptions as agents consume computing resources much faster than ordinary chat use. InfoWorld likewise described the move as part of a broader shift toward metered pricing for AI agents. InfoWorld cited criticism from developers including Yadesh Salvi, a senior data scientist, and Advait Patel, a senior site reliability engineer at Broadcom. (infoworld.com) Patel said the shift matters for developers who built side projects and personal automations around flat Pro or Max plans, while Salvi said the new monthly allotment would not cover a day of serious work for some users. (axios.com) ### What still stays inside the regular Claude subscription? Anthropic’s pricing page still markets Pro and Max as consumer-style subscriptions for chat, coding and productivity features on Claude’s web, mobile and desktop surfaces. The page says Pro includes Claude Code and Max provides 5x or 20x more usage than Pro, while also warning that usage limits apply. The distinction after June 15 is that direct user interaction inside Claude remains part of the subscription structure, while automated agent runs through programmatic channels will draw down a separate monthly credit pool first and then convert to metered charges, according to InfoWorld’s account of Anthropic’s policy. (infoworld.com) That leaves developers to track two kinds of consumption: ordinary Claude plan usage and agent-specific credit usage. (claude.com) ### What happens next on June 15? June 15 is the date InfoWorld said Anthropic will begin separating programmatic Claude usage from standard chat limits and applying the new monthly credit pools. Anthropic’s pricing page remains the public reference point for plan prices, while developers using the Claude Agent SDK, GitHub Actions and outside agent frameworks are the customers most directly affected by the change. (infoworld.com)