Tiny Spoon

Big AI news, in small bites

STRATEGYAnthropic
PROGRAMMATIC USE METERED$$$$$JUN15 ANTHROPIC JUN 15

Anthropic is splitting Claude subscriptions. Starting June 15, programmatic usage - Agent SDK, claude -p, GitHub Actions, third-party agents - moves to a separate metered credit pool. Pro gets $20/mo, Max 5x gets $100, Max 20x gets $200. Billed at full API rates. Credits don't roll over.

Anthropic just metered the Claude agent ecosystem. Interactive use stays flat-fee. Anything programmatic now burns API-rate tokens against a separate monthly bucket.

The split is precise. Agent SDK, claude -p non-interactive mode, Claude Code GitHub Actions, and third-party apps authenticated via Agent SDK all move to the new credit. Pro ($20), Max 5x ($100), Max 20x ($200). Sonnet 4.6 at $3/$15 per million in/out tokens. Opus 4.7 at $5/$25. The credit resets monthly and does not roll over - use it or lose it. Claude.ai web, desktop, mobile, Cowork, and the interactive Claude Code CLI in your terminal are all unchanged.

For PMs: if your team's CI/CD relies on Claude Code GitHub Actions, model your June-15 burn rate now. For execs: this is the second usage-based reprice in dev-AI in two weeks - GitHub Copilot was June 1. The category is repricing. For founders building on Agent SDK: budget at full API rates, not subscription rates.

▾ full brief & sources

Why this matters

  • Subscription unit economics for high-usage agents never worked. Anthropic is calling it.
  • Mirrors GitHub Copilot's June 1 move - the second major dev-AI reprice in two weeks. The category is repricing in real time.
  • Forces a budgeting conversation inside every team running Claude in CI/CD - the line item changes from "Pro seat" to "agent compute."

🔍 What happened

  • Effective June 15, 2026. Anthropic splits Claude subscription credits into two pools.
  • What moves to the new metered pool: Agent SDK, claude -p (non-interactive Claude Code), Claude Code GitHub Actions, third-party apps authenticating via the Agent SDK.
  • Credit amounts by plan: $20 for Pro, $100 for Max 5x, $200 for Max 20x.
  • Billing: full API rates. Sonnet 4.6 at $3 input / $15 output per million tokens. Opus 4.7 at $5 / $25.
  • Credits do NOT roll over - monthly reset, use it or lose it.
  • What stays unchanged: interactive Claude Code in the terminal, Claude.ai web/desktop/mobile, Claude Cowork.
  • Users must claim the new credit through their Claude account before June 15.

💬 Smart takes

  • The New Stack: Anthropic splits billing again - Agent SDK gets separate credit pools. The framing inside the dev community is "again" because this is the third pricing shift in 90 days.
  • Vaught AI: Anthropic Just Put Claude Agents on a Meter - the operator's read.
  • Zed Blog: What Anthropic's New Claude Billing Means for Zed Users - downstream tools are writing their own user-facing explainers.
  • Skeptic, VentureBeat: Anthropic reinstates third-party agent usage on Claude subscriptions - with a catch. The reinstating-with-a-catch frame says the dev community pushed back.

🧭 Where this goes

  1. Cursor and Codex CLI ship credit-pool models within 30 days, or hold flat-fee and bleed power-user money.
  2. Claude Code GitHub Actions usage drops 40-60% in the first month as teams audit their burn.
  3. Local-model setups (DeepSeek R1, Qwen 3.7) get a wave of dev adoption from refuseniks.
  4. Anthropic's API revenue mix shifts toward enterprise contracts and away from consumer subscriptions - expect that in the S-1.
  5. Third-party Agent SDK apps build credit-aware UX (per-call cost displays, monthly burn dashboards) within Q3.

🎯 Implication

  • For PMs: if your product calls Claude via Agent SDK, model June-15 cost at full API rates. The subscription was the subsidy.
  • For execs: dev-AI is now a metered cost center. Treat it like cloud spend - allocate budgets, monitor burn, set caps.
  • For founders: Agent SDK apps that assumed flat-fee unit economics need to reprice. Most haven't yet.