CC Probeline

cc-probeline

A status line for Claude Code that prices every turn — yours, your subagents', and every cache rebuild — reading the session log already on your disk. No account, no API key, no telemetry.

Session totals come from Claude Code itself; the per-turn breakdown is computed here from token counts against a public price table, and labelled as an estimate throughout.

It makes no network calls while it runs. Two optional, opt-out background actions exist: a once-a-day check of one public file on GitHub (current prices and the latest version), and a refresh of Claude Code's own usage cache — its /usage screen run headlessly, once when a session starts and rarely after that, and only while your plan actually has a model-scoped weekly limit or paid extra usage to keep current. Neither sends anything about your session. Privacy policy →

Spend your limits on purpose, instead of paying for inefficiency you can't see.

Install:

Windows (Scoop) and the Claude Code plugin marketplace: all install options →. Every release ships with SHA256 checksums and signed SLSA build provenance.

What the probe pulls out

Most status lines count things — tokens, turns, running agents. The probe prices them. Everything below comes out of your session's local log: data Claude Code has, but never shows you.

- Every turn, priced — not one opaque session total: a live table where each step lands with its own cost.
- What your subagents spend — subagent work is invisible while it runs. The probe puts each agent on the bill, live, next to your own turns.
- Cache rebuilds, in dollars — idle past the TTL (60 min for the orchestrator, 5 for subagents), and your next turn quietly rewrites the whole cache. The probe ages it live (⏱ 60m → 0m) and prices the rebuild when it hits.
- Extra usage…