Claude's free ride ends June 15.
4 June 2026 · Issue #8
Anthropic just changed how billing works. Here's what it means for your workflows.
Got an email from my accountant this week asking if I'd "reviewed my AI spend". I hadn't. Apparently that's a thing now. Welcome to the metered era.
The bill just changed
Starting 15 June, Anthropic is separating how it charges for programmatic usage. If you're running any Claude workflows through the Agent SDK, GitHub Actions, or third-party tools like OpenClaw, those will no longer draw from your standard subscription. They'll pull from a dedicated monthly credit pool instead, billed at API rates.
What does that mean in actual money? Pro users get £20 in credits. Max 5x users get £100. Max 20x gets £200. If your automated workflows stay within that, you're fine. If they don't, you pay overage.
The framing from Anthropic is that this "separates" the two use cases rather than cutting anything. That's technically accurate. What it also does is end the quiet subsidy that made running agent workflows on a £20 plan genuinely good value. As one analyst put it to InfoWorld: treat your Claude usage the way you treat AWS. Know your cost per workflow.
For most SMBs running one or two lightweight automations, the credit allocation will cover it comfortably. But if you've built anything more ambitious, check your usage before June 15. The /usage command in Claude Code now shows a per-category breakdown of what's burning your limits.
ChatGPT now has ads. Sit with that.
OpenAI launched its ChatGPT Ads Manager beta on 5 May. Ads inside your AI assistant. Clearly labelled, OpenAI says, displayed at the bottom of responses, won't influence the answer. That's the same thing every ad-supported platform has said at launch. Google said it in 2000. Facebook said it in 2012.
Anthropic's response, published on its news page, is worth reading in full: "Users shouldn't have to second-guess whether an AI is genuinely helping them or subtly steering the conversation towards something monetizable." For a business tool, that distinction matters more than it does for a consumer app. You're asking Claude about pricing strategy, supplier negotiations, customer complaints. The idea that a sponsored result might be adjacent to that answer should feel wrong.
Anthropic can afford to take this position because roughly 80% of its revenue comes from enterprise contracts, not individual subscriptions. The ad-free pledge isn't just principles. It's a product strategy built on a different business model. Which doesn't mean it'll hold forever. But for now, you're not the product.
Three things to try this week
Two of these cost nothing. One requires a paid plan. All three are genuinely useful right now.
Free plan got better — File creation, connectors, and skills are now available on the free Claude plan. If you've been holding off on recommending Claude to a colleague because of the cost, that barrier just dropped. Have them connect Google Drive and try a basic research task.
Manage your Meta ads in plain English — Meta released an official MCP giving Claude access to 29 tools for managing Facebook and Instagram ad accounts, free during beta. If you run paid social, this is worth 20 minutes of your time. You can pull performance data, adjust budgets, and draft copy without touching Ads Manager.
Check your usage breakdown — Before June 15, open Claude Code and run /usage. It now shows a per-category breakdown of what's consuming your limits across skills, subagents, and plugins. Worth knowing before the billing change kicks in.
From the timeline
The June 15 billing change landed quietly. The reaction was not quiet.
The automation that quietly costs more
You set up a Claude workflow three months ago. It pulls from a Google Sheet every Monday, drafts a status update for your three biggest clients, and drops it into a shared folder. Took you an afternoon to build. Saves you about two hours a week. You haven't thought about it since.
On 16 June, you get an email. You've exceeded your credit allocation. The workflow ran fine. You just didn't know it was drawing from a separate pool that didn't exist a week ago.
This is the scenario the billing change creates for anyone running background automations. Not a disaster. But a surprise if you haven't checked. The fix is simple: run /usage in Claude Code before the 15th, look at the "skills and subagents" category, and estimate whether your monthly workflows stay within your credit tier. Pro users get £20 worth. For most lightweight automations, that's plenty. For anything hitting APIs repeatedly or running multi-step agents, do the maths now rather than after the invoice.
The InfoWorld piece on the change is worth bookmarking. It explains the new credit tiers clearly, and the advice from Doozer AI's co-founder is direct: optimise for the token, not the subsidy.
Re-reading the original Google IPO letter from 2004. The one where they promise not to let ad revenue compromise search quality. Lovely document. Worth knowing how that one ended.