June 15 Cliff Exposure — from your real logs

On 2026-06-15, programmatic Claude Code usage (claude -p, the Agent SDK, anything in CI/cron) moves into a separate paid pool (Pool 2) with a fixed monthly budget per plan. Interactive claude sessions are unchanged. Most "am I affected?" tools make you estimate. This one reads the entrypoint your logs actually recorded and splits your usage by it. (Want a 20-second version with no logs? Use the questionnaire check instead — this page is the precise, real-data companion. 日本語版.)

🔒 100% in your browser. Your logs are parsed locally with JavaScript and never uploaded — there is no server. Disconnect from the network and it still works.

1. Pick your session logs

They live in ~/.claude/projects/ (each project is a folder of .jsonl files). Select that folder, or drag in any .jsonl files.

or drag & drop .jsonl files here

2. How much of you is programmatic?

of your recent activity ran as a programmatic invocation (the part that moves to Pool 2)
invocation (entrypoint)pool from Jun 15sessionsactivity (msgs)

Your heaviest programmatic sessions (if any)

sessionentrypointactivity (msgs)working dir

3. What this means for you

Honest limits — read these before acting:
  • This classifies whole sessions by their recorded entrypoint — the dimension June 15 actually splits on. cli = interactive (Pool 1). sdk-cli and anything non-cli = programmatic (Pool 2).
  • Headless claude -p may be recorded as cli in some versions. If you run Claude in scripts, cron, or CI and it isn't showing up below, this view understates your Pool 2 exposure — trust your own knowledge of your automation over this number.
  • "Activity" is the count of assistant turns — a volume proxy, not exact tokens or dollars. A few heavy programmatic sessions matter more than many tiny ones.
  • It does not predict how subagents, hooks, or MCP inside an interactive session are billed. Anthropic has not defined that split; anyone asserting it is guessing. This tool stays on the part that's actually defined.

You know your exposure — now pick a response path

If a real share of your usage is programmatic, the question becomes which of four paths fits you (fold into interactive / optimize claude -p / move to the API / mix). The June 15 cliff survival guide (¥800, JP) walks each path with its cost, limits, and steps, plus the single metric M and an 18-day timeline. The first two chapters are free.

To cut the underlying consumption itself: Token Book — JP (¥2,500) · EN (name your price).

Free guardrails against runaway programmatic cost: cc-safe-setup — npx cc-safe-setup.

Runs entirely client-side. No data leaves your machine. Part of cc-safe-setup.