Anthropic Splits Agent SDK Off the Subscription — Credits at API Rates from June 15

Subscription simplicity subsidized agentic automation until it didn’t. Anthropic’s split is the admission that SDK/CI workloads are API economics — chat stays in the plan bucket; agents get their own meter.

Anthropic (subscriber email, May 13; effective June 15, 2026): usage of the Claude Agent SDK, claude -p, Claude Code GitHub Actions, and third-party tools using the Agent SDK will no longer draw from regular Claude subscription rate limits. Instead: a dedicated monthly credit billed at standard API rates, sized to the plan — $20 (Pro), $100 (Max 5x), $200 (Max 20x), with Team/Enterprise equivalents. Unused credits do not roll over.

What changes

Plan	Agent SDK monthly credit
Pro ($20/mo)	$20
Max 5x ($100/mo)	$100
Max 20x ($200/mo)	$200

Interactive use (claude.ai chat, standard Claude Code) stays on subscription limits. Programmatic/agentic workloads move to the credit bucket at API list prices.

Claims vs checks

Change details come from subscriber email / support docs as reported in May 13–14 developer coverage and gists — not a flashy marketing launch. Cost impact for high-volume agent users is the real story; Anthropic frames it as aligning costs with compute-intensive agent workloads. Same week, OpenAI courted switchers with free Codex months.

Limits

• Exact Team/Enterprise credit tables follow tier equivalents — confirm in account docs.
• No rollover raises use-it-or-lose-it pressure.
• Third-party tool coverage depends on whether they call the Agent SDK path.

Sources

• Anthropic subscriber email and support documentation (May 13, 2026) as compiled in developer discussions.
• Contemporaneous reports (May 13–14) on email contents, effective date, and credit tiers.
• Anthropic platform docs on Agent SDK usage.