Claude Agent SDK Billing Splits on June 15: What the New Credit Pool Means for Your Plan
From June 15, Claude Agent SDK, claude -p, GitHub Actions, and third-party agents draw from a separate monthly credit, not your subscription quota.
On June 15, Anthropic changes how programmatic Claude usage is billed for Pro, Max, Team, and Enterprise subscribers. If you use the Claude Agent SDK, the claude -p command, Claude Code GitHub Actions, or any third-party tool that hooks into your Claude subscription, those requests stop drawing from your plan’s standard usage quota and start drawing from a new, separate monthly credit pool instead.
This is a meaningful structural change if you run any kind of automation on a paid Claude plan. Here is exactly what is happening and what you need to do about it.
The Two Buckets
Before June 15, all Claude usage on a paid plan shared the same pool of rate-limited capacity. Your chat sessions, your terminal coding sessions, and your background agents all competed for the same quota. That model worked when programmatic use was light. It stopped working well when autonomous agents started generating thousands of requests per day against a flat subscription fee.
From June 15, usage splits into two clearly separated buckets.
Bucket one (unchanged): Interactive use. This covers Claude.ai chat, Claude Code running in your terminal or IDE, and Claude Cowork sessions. Nothing changes here.
Bucket two (new): Programmatic use. This covers the Claude Agent SDK (Python and TypeScript), the claude -p non-interactive CLI command, Claude Code GitHub Actions, and third-party apps that authenticate using your Claude subscription. These now draw from a dedicated monthly credit denominated in dollars.
The two buckets do not overlap. If your Agent SDK credit runs out, your interactive Claude sessions are completely unaffected. And interactive usage will never accidentally eat into your programmatic credit.
How Much Credit Do You Get?
Credit amounts are pegged to your plan’s monthly fee:
| Plan | Monthly Agent SDK Credit |
|---|---|
| Pro | $20 |
| Max 5x | $100 |
| Max 20x | $200 |
| Team Standard | $20 per seat |
| Team Premium | $100 per seat |
| Enterprise | Varies by seat type |
Credits are per user, not pooled. A team with five Max 5x seats has five separate $100 credits, not one shared $500 pool. Enterprise Standard seats are not eligible to claim the credit at all, which is worth flagging if you manage a plan where engineers are running automation.
Credits reset with your billing cycle. There is no rollover. Unused credit from one month does not carry forward.
One important catch: you have to claim the credit once before it activates. After that first claim, it refreshes automatically each month.
What Happens When the Credit Runs Out?
This depends on a setting Anthropic calls “usage credits.” Think of it as an overflow toggle.
When usage credits are enabled on your account, Agent SDK usage continues after the monthly credit is exhausted and bills at standard API rates. When usage credits are disabled, requests are simply rejected until your credit resets the following month.
To put that in concrete terms: if you are on a Pro plan with $20 of monthly credit and you run a GitHub Action that triages issues daily, you might burn through that credit in under two weeks. Whether your automation pauses at that point or keeps going (at your expense) is entirely down to whether you have that overflow toggle enabled.
At Sonnet 4.5 API rates ($3 per million input tokens, $15 per million output tokens), a typical issue-triage run consuming 15,000 input tokens and 2,000 output tokens costs roughly $0.075. Twenty runs a day is $1.50. Over 30 days, that is $45. The $20 Pro credit covers about the first 13 days of that workflow.
What You Need to Do Before June 15
If you only use Claude interactively: Nothing changes. You do not need to take any action.
If you use the Agent SDK, claude -p, GitHub Actions, or third-party agents:
- Go to your account settings and claim the monthly Agent SDK credit. It will not activate automatically.
- Decide whether you want usage credits (overflow billing) enabled. If you do not want surprise charges, leave it off and accept that automation pauses when the credit is gone.
- Audit your workflows against the credit amount for your plan. If $20 does not cover your Pro-plan automation, either upgrade your plan or budget for overflow charges.
One More Thing Happening on June 15
At the same time as the billing change, Anthropic is retiring claude-sonnet-4-20250514 and claude-opus-4-20250514 from the API. If any of your production agents or scripts still reference those model IDs on June 16, every request will return an error. Migrate to the current model versions before the cutover date.
The Bigger Picture
Anthropic’s framing is that subscribers are receiving a “free credit” for programmatic use. The more accurate framing is that a subsidy that previously existed inside the flat subscription fee is now being made explicit and capped. Autonomous agents are expensive to run at scale, and flat-rate subscriptions were not really priced to support unlimited agentic workloads.
For most individuals using Claude interactively, this is a non-event. For developers and teams running background automation on subscription plans, it is the most significant billing change since Claude Code launched. Check your workflows, claim your credit, and set your overflow preference before the 15th.
Full details are in Anthropic’s official Help Center article on Agent SDK credits and the Claude Platform release notes.