For people who live in Claude Code

See what your sessions cost. Then cut it.

ReadyClaw reads the transcripts Claude Code already keeps on your machine, prices every turn at list rates, and finds the settings and habits that quietly burn tokens: the pinned old subagent model, the 900k-token context nobody compacted, the lockfile that got read into memory four times.

readyclaw · last 30 days · 3,542 transcript files

API-equivalent spend   $13,508
Tokens seen            22.6B (cache hit 98%)

[HIGH] 62% of turns re-sent more than 200k tokens of context   (~$6,465)
[HIGH] Subagents are pinned to claude-sonnet-4-6, a superseded model   (~$644)
[MEDIUM] 113 turns resumed a session after an hour and rewrote the whole context   (~$504)

Headroom: up to $10,679 of $13,508 (79%) is recoverable

Local first

One command reads ~/.claude/projects and prints the report. Nothing leaves your machine unless you sync, and a sync carries only daily totals, session totals and findings. Never prompts, never code.

Headroom, in dollars

Every finding comes with a measured estimate and a concrete fix. The Headroom command applies the safe ones to your settings with a backup, one confirmation at a time.

Leaks and anti-patterns

Every credential that slipped into a prompt, command or tool output is listed by kind with how to rotate it, never the value itself. The guard blocks the next one before it is sent, and the habits behind them are counted so you can stop them.

Honest numbers

Duplicate streaming lines are collapsed, cache reads are priced at their real rate per model, and overlapping findings are counted once. Estimates say when they are upper bounds.

Desktop app · macOS and Windows

Catch context bloat while it happens

A tray item shows the largest live session's context and what each turn now costs. The first time a session passes 200k, 400k and 700k tokens you get a notification. Turn on the guard and Claude Code itself is told when a session has grown past the threshold, so it offers /compact or a fresh session before continuing. Settings fixes apply with one click, with a backup.

Free to scan and watch; one-click fixes and the guard are on Pro. macOS builds are signed; the Windows build is unsigned for now, so SmartScreen asks once.

Pricing

Pays for itself on the first finding.

Free

$0/mo

The CLI, and a week of history in the dashboard.

  • readyclaw scan, unlimited, local only
  • 7-day dashboard window
  • Per-model and per-project spend
  • CSV export
  • 1 machine
Start free

Pro

Most useful
$9/mo

Everything you spend, every leak that wastes it.

  • Unlimited history
  • Recommendations with dollar estimates
  • Headroom: guided settings fixes
  • Per-session drill-down
  • 3 machines
Get Pro

Team

$29/mo

Pro for up to five people, with the spend and the leaks rolled up.

  • Everything in Pro, for every member
  • 5 seats, invited by email
  • Per-person spend, sessions and guard status
  • Every open leaked credential across the team
  • Guard policy: require it on every machine
Get Team

Yearly billing is two months free. Cancel any time from the billing portal.