If you searched for “Claude pricing” or “Claude free tier” recently and found conflicting answers, you are not alone — search interest in Claude’s plans exploded this June. The reason: Anthropic announced a significant restructuring of how Claude subscriptions are billed, set it to take effect on June 15, 2026 — and then, on the day itself, paused the whole thing. Here is exactly what was announced, what actually changed (spoiler: for now, nothing), and what Claude really costs in July 2026.
The short version
- May 14, 2026: Anthropic announced it would split subscription billing into two pools — one for Claude’s own apps, one for Agent SDK and third-party tools — with a new monthly “Agent SDK credit” ($20 on Pro, $100 on Max 5x, $200 on Max 20x), effective June 15.
- June 15, 2026: Anthropic confirmed the change is not going ahead for now — the pause was announced in its Help Center and in notes to subscribers.
- Today: Agent SDK, claude-p and third-party apps still draw from your normal subscription limits, exactly as before. No separate credit exists.
- Why it matters: the episode shows how expensive agentic AI usage has become — and why every AI vendor is rethinking how to bill for it.
What Anthropic announced in May
On May 14, Anthropic told subscribers that from June 15 its Claude subscriptions would be split into two usage pools. Using Claude through Anthropic’s own surfaces — the web app, mobile apps, and Claude Code — would continue to draw from your plan’s normal limits. But usage flowing through the Claude Agent SDK and third-party applications built on it — the growing ecosystem of editors, agents and tools that connect to a Claude subscription — would move to a separate monthly Agent SDK credit: $20 worth on Pro, $100 on Max 5x and $200 on Max 20x.
The logic was not hard to read. Agentic workloads are token-hungry — an autonomous coding agent can burn through in an hour what a chat user consumes in a month. A flat subscription that treats both the same is a pricing model under strain. Separating the pools would have capped how much third-party agent usage a subscription could carry.
The backlash, and the pause
The reaction was swift. Developers who had built workflows on third-party tools — and the makers of those tools, including code editors that let users bring their Claude subscription — warned the change would effectively shrink what subscribers were paying for. Commentators framed it as another quiet reduction in Claude Pro’s value, and community forums filled with cancellation threats and confusion about what would still be included.
Then came the reversal: on June 15, the day the change was due, Anthropic confirmed that the Agent SDK credit was no longer happening — for now. In its Help Center and in a note to subscribers, the company said ACP usage, claude-p, the Claude Agent SDK and third-party apps built on it would continue to work with subscriptions exactly as before, with unchanged limits and no separate credit to claim. Anthropic left the door open to revisiting the model later, which is worth remembering: the underlying economics have not gone away.
What Claude actually costs right now (July 2026)
With the noise cleared, here is the practical picture of Claude’s plans as of July 2026 — unchanged by the paused overhaul:
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Claude web + apps with daily usage limits — enough for casual use, resets regularly |
| Pro | ~$20/mo | Much higher limits, priority access, Claude Code and projects — the standard individual plan |
| Max 5x | ~$100/mo | Roughly 5× Pro limits for heavy users and agent workflows |
| Max 20x | ~$200/mo | Around twenty times the Pro limits, built for power users who run agents all day |
| Team / Enterprise | per seat | Collaboration, admin controls, higher limits |
| API | per token | Pay-as-you-go per million tokens; the route for building products |
Two practical notes. First, the free tier remains real — the surge of searches for “Claude free tier” this June was largely people checking whether it survived the changes; it did. Second, subscriptions and the API remain entirely separate: a Pro plan does not include API credits, and API usage is priced per model — you can compare what every Claude model (and its rivals) costs per million tokens in our AI models database, or estimate a real monthly bill with the AI API cost calculator.
The bigger story: agents broke the subscription model
This episode is bigger than one company’s billing page. The industry’s flat-rate subscription model was designed for chat — a human typing, reading, thinking, typing again. Autonomous agents changed the equation: tools like Claude Code and the wave of agentic coding assistants can run for hours, making thousands of model calls without a human in the loop. Every vendor is quietly wrestling with the same question: how do you price unlimited-feeling access to something whose marginal cost scales with how autonomous it gets?
Expect more experiments like this across the industry — separate agent pools, usage-based tiers, compute credits. Anthropic paused this one, but the pressure that produced it is structural. For cost-sensitive workloads, this is also why open-weight alternatives keep gaining ground: our open-vs-closed cost study found open models average roughly 16× cheaper per token, and models like DeepSeek V4 exist precisely for the workloads where per-token economics dominate.
What subscribers should do now
Practical guidance, in order of effort. If you only use Claude’s own apps and Claude Code, do nothing — none of this touched you, and it would not have even if the change had shipped. If you rely on third-party tools connected to your subscription, take five minutes to note which ones matter to you and follow Anthropic’s Help Center announcements — a revived version of the split would affect them first, and vendors like your editor’s maker will typically post migration guidance quickly. If you are running heavy agentic workloads on a Pro plan, treat this as the warning shot it is: measure your actual token consumption now, so that if usage-based pricing arrives you can compare a Max plan, direct API pricing, or a cheaper model tier with real numbers instead of guesses. Our cost calculator makes that comparison trivial — and knowing your numbers before a pricing change is the difference between a calm migration and a panicked one.
Frequently asked questions
Did Claude’s pricing change in June 2026? No. A billing restructuring (the Agent SDK credit) was announced for June 15 but paused before taking effect. Plans and limits work as they did before.
Is Claude still free to use? Yes — the free tier remains, with daily usage limits. Pro (~$20/mo) lifts them substantially.
What was the “Agent SDK credit”? A proposed separate monthly allowance ($20/$100/$200 by plan) for usage through the Agent SDK and third-party apps. It never took effect.
Could Anthropic bring the change back? Yes — the company paused rather than cancelled the idea, and the economic pressure from agentic usage is industry-wide. Watch official announcements.
Do Claude subscriptions include API access? No. The API is billed separately per token; subscriptions cover Claude’s apps and Claude Code.
The bottom line
Nothing about your Claude bill changed this June — but the episode was a preview of where AI pricing is heading. Agents consume tokens at a scale flat subscriptions were never designed for, and every vendor will eventually face the question Anthropic just punted on. If you rely on subscription-powered third-party tools, enjoy the reprieve, keep an eye on the official help pages, and know your fallback options — because in 2026, the smartest AI budget is one that has already priced the alternatives.
Compiled July 2, 2026 from Anthropic’s Help Center notices, subscriber communications and industry reporting. Prices are approximate and may vary by region; check claude.com/pricing for current figures.