Unlimited AI coding subscriptions are starting to meet the meter.
Anthropic said paid Claude subscribers will get a separate monthly credit pool for agent tools beginning June 15, covering Claude Agent SDK usage, the claude -p command, GitHub Actions, and third-party frameworks built on the Agent SDK. The change means automated workloads will no longer draw from the same bucket as standard Claude chat usage.
For developers using Claude plans to run agents, scripts, and CI workflows, the update turns a predictable subscription perk into something closer to cloud billing: useful, powerful, and easier to overspend if no one is watching the dial.
What’s changing
According to Anthropic, the monthly Agent SDK credit will cover Claude Agent SDK usage, the claude -p command, and third-party apps built on the Agent SDK. The credit is only for Agent SDK usage and refreshes with your billing cycle.
Pricing will depend on a user’s existing Claude subscription, with Pro users receiving $20 in credits, Max 5x users $100, and Max 20x users $200.
“When your monthly credit runs out, additional Agent SDK usage flows to extra usage at standard API rates — but only if you’ve enabled extra usage,’’ Anthropic said. “If extra usage isn’t enabled, Agent SDK requests stop until your credit refreshes.”
A sign of the times?
The move paves the way for rival OpenAI to lure power users to Codex. It also underscores that unlimited AI subscriptions may not survive the agent era, given software’s ability to deplete computing resources more quickly than humans.
Anthropic’s decision is part of a larger industry shift toward metered pricing for AI agents, which means developers and enterprises will have to rethink the financial feasibility of large-scale automation workloads.
Developers will likely have to pivot
Another April post about the Claude subscription change effectively hammered home the point that developers would be forced to use external agent frameworks either by purchasing additional usage bundles or by switching to direct API access.
Previously, the same subscription pool covered programmatic workloads and interactive Claude usage, allowing developers to use higher-tier Claude plans for autonomous agents, scripts, CI pipelines, and other automated workflows — in addition to chat and coding assistance.
That arrangement made Claude subscriptions especially appealing to developers running persistent agent workloads, because usage through tools like OpenClaw or Agent SDK was generally absorbed into the subscription’s usage limits rather than charged separately at API pricing.
Strong reaction to the change
It likely comes as no surprise that the policy change has alarmed developers, many of whom claim it erodes one of Claude’s biggest draws for agentic workflows: the ability to have relatively predictable subscription pricing while running large-scale automations.
After Claude Code product manager Noah Zweben’s X post about the new rules, developers didn’t hold back in their responses.
For example, one wrote: “really crap! as a solo dev I was using sdk to help my workflow. guess I’ll be switching effort towards my codex sub, which was my back up until now. bye (if you have no idea 200$ of credit is a couple of hours of coding even “interactive.”)”
And another reacted with: “Guys are you serious with that communication bull****? Framing this as a positive step? I get why you need to save costs. Just say it and stop fooling your user base. Holy moly.”
Why metered pricing is gaining traction
The shift away from unlimited plans reflects a practical reality: autonomous agents consume far more compute than human users because they operate continuously and generate high-volume requests.
- Unlike human users, agents can run continuously, execute tests, browse the web, and make repeated model calls without natural stopping points.
- That level of activity can drive usage sharply higher and produce significant AI bills for businesses.
- Some companies, including ServiceNow and Uber, have reportedly already exhausted their annual AI token budgets, according to The Information.
The broader takeaway is that everyone is in the same boat. As AI agents become more capable and more widely used, metered pricing may increasingly replace unlimited access across the industry.
Also read: Anthropic could reach a $1 trillion valuation as investor interest in Claude, enterprise AI, and agentic tools continues to grow.