See where your Claude Code spend actually goes.

You're using Claude Code every day with no feedback loop — costs, errors, and retry loops fly by invisible. Qualito measures every session so you can see patterns, cut waste, and prove your setup is getting better.

Get Started →
uvx qualito setup
View on GitHub
Terminal
$ qualito setup
  Scanning ~/.claude/projects/ ...
  Found 8 projects · 312 sessions · 30 days
  Importing...
  ████████████████████████████ 312/312
  Last 30 days
  ┌─────────────┬──────┬──────────┬─────────┐
  │ Workspace   Runs Duration Cost   
  ├─────────────┼──────┼──────────┼─────────┤
  │ my-app         89   14h 12m $342.80
  │ api-service     67    9h 48m $218.40
  │ frontend        54    7h 32m $186.20
  └─────────────┴──────┴──────────┴─────────┘
  Total: $847.40 · 312 sessions · 6 over $50
  Dashboard ready → app.qualito.ai
Works with Claude Code
on PyPI
MIT licensed CLI
Local-first · your data stays on your machine

Within 10 seconds of connecting,
you'll see something you didn't know about yourself.

Three real patterns Qualito surfaces on day one. No setup, no tagging, no configuration. Just your own data, framed in a way you haven't seen it before.

$847/30d
6 sessions cost 40% of your monthly spend.
Your biggest sessions hide in plain sight. Qualito ranks them by cost so you can click in and see what happened.
23% errors
1 in 4 Bash calls in your setup fails.
Tool error rates broken down per command. Fix a broken pre-commit hook once and stop paying the tax every session.
47 retries
The same error fired 12 times in a row.
Spot the retry loops. See the exact session where Claude got stuck, and the error it couldn't get past.

Nobody sold you a problem. The data did.

Start transparent. Get scientific.

Each layer builds on the previous. Most users live in layers 1–2 and find enormous value there. Pro is for when you want to prove things.

  1. 01

    Transparent

    See what happened.

    Cost, duration, tools, files — every session, from day one. Zero friction, zero workflow change.

  2. 02

    Observable

    See what's changing.

    Trends over time, anomalies, per-workspace patterns. Spot the session that cost 10× your average.

  3. 03

    Measurable

    Score your quality.

    Tag three sessions and DQI unlocks — a composite quality score tuned to your own signal.

  4. 04

    Scientific Pro

    Prove what works.

    A/B observational experiments. Wilcoxon + Bayesian validation. Real statistics, not vibes.

  5. 05

    Self-improving Pro

    Get better automatically.

    AI-designed experiments, auto-generated fix proposals, template extraction from your best-performing sessions.

Four kinds of reader, one dataset.

Individual developers

See your usage. Understand your patterns. Stop guessing at your AI costs.

Tech leads

Find what works across your team's sessions. Improve the practice, not the vibes.

Systems builders Pro

A/B test prompting strategies with statistical rigor. Prove the wins before you roll them out.

Engineering directors Pro

Cost attribution and compliance for AI-assisted development. Point at the data, not the vibes.

Install. Import. See insights.

Qualito reads the JSONL files Claude Code already writes. No workflow change, no API tokens to rotate, no instrumentation to wire up.

1

Install

One command. Python 3.11+, uv, done.

uvx qualito setup
2

Connect

Register, paste the one-time setup token. Qualito imports the last 30 days of sessions locally.

→ app.qualito.ai/login
3

See

Cost, duration, tools, and patterns across every project — live from minute one.

open the dashboard

Layer 1 and Layer 2 — what you get for free.

Session classification

Interactive, delegated, or SDK

Every session labelled by how it ran so you can compare costs and patterns apples-to-apples.

Cost breakdown

Workspace, model, day

Slice your spend by project, by Claude model, by day. Find the expensive outliers without a spreadsheet.

Tool usage + error rates

Patterns, not totals

Which tools you use, how often, and where they fail. Surface the broken commands you've been paying a tax on.

Run detail

Transcript, tool calls, files

Click any session for the full timeline: conversation, every tool call, files touched, and the running cost.

Incident detection

Spikes, loops, regressions

Qualito flags error spikes and cost anomalies automatically. You get alerted, not surprised at the end of the month.

Privacy-first sync

Per-workspace control

Default to metadata-only. Opt into full content per project. Every sync is scanned for secrets before anything leaves your machine.

Prove your process is getting better.

Observational A/B experiments on your own sessions. Wilcoxon signed-rank for significance. Bayesian posterior for the size of the effect. Not a gut feeling — a verdict.

A: Clear task up front WINNER
0.82
Sessions 24
Avg cost $1.40
Completed 91%
vs
B: Minimal prompts
0.61
Sessions 24
Avg cost $3.20
Completed 67%

Significance: p = 0.003 · effect size: +34%

Adopting strategy A would save $43/month at your current volume.

Observational — no workflow change
Runs on metadata — content stays put
Verdicts backed by real statistics
See everything in Pro →

Free today. Pro when you want to prove things.

Free
Current
$0 forever
  • Unlimited runs
  • Cloud dashboard
  • Incident detection
  • Recommendations
  • SLO defaults
  • CLI + MCP server
  • Local-first, metadata-only sync by default
Get Started →
Pro
Waitlist
$– / month — pricing finalising.
  • Everything in Free, plus:
  • DQI scoring on tagged runs
  • A/B experiments · Wilcoxon + Bayesian
  • LLM-designed fix experiments
  • Custom SLO targets per workspace
  • Recommendation detail pages
  • Priority support
See Pro features →

Common questions.

Your next Claude Code session is already measurable.

See it.

Get Started →
uvx qualito setup

Free tier · no credit card · local-first by default.