Qualito

Quality metrics for AI-assisted development. Know if your Claude Code sessions are worth the cost.

What it does

Qualito scores every Claude Code session on multiple quality dimensions (error rate, tool diversity, cache utilization, completion with work) and rolls them into a single Score (0-100). It tracks cost, detects incidents (error spikes, cost anomalies, quality drops), and surfaces actionable recommendations.

• Read existing Claude Code sessions — no behavior change needed
• Score each session across multiple dimensions (0-100)
• Track costs, detect incidents, monitor quality trends
• Set SLOs and get alerts when quality drops
• Compare runs before/after a change (observational experiments)

Quick Start

See docs/getting-started.md for the full walkthrough.

Short version:

1. Sign up at https://app.qualito.ai/login
2. Copy the setup command shown on the onboarding page (includes a one-time token)
3. Paste into your terminal:

   uvx qualito@latest setup st_setup_<your_token>
   

4. Refresh the dashboard → your sessions are live

Cloud Dashboard

Free tier: 3 workspaces synced. Pro ($29/mo, coming soon): unlimited workspaces + LLM-generated root-cause + report summaries.

CLI Commands

View your data

Qualito is CLI-first. Your data lives locally and is accessed through commands:

qualito status       # local + cloud breakdown with per-workspace tokens + cost
qualito costs        # detailed spend analysis
qualito privacy      # per-workspace sync privacy settings
qualito audit list   # flagged runs needing review

For a web UI with charts and history, sync to the cloud and view at https://app.qualito.ai.

Privacy

Qualito is local-first. Your session data lives on your machine, and you control what syncs to the cloud on a per-workspace basis.

• Default: metadata only (counts, durations, types, scores)
• Opt in per workspace to sync full content for a richer dashboard
• Every sync scanned for secrets (AWS keys, API tokens, passwords, etc.)
• Server enforces your stated settings — defense in depth

See docs/privacy.md for the full field-by-field breakdown.

MCP Server

Use Qualito inline in your editor via MCP:

{
  "mcpServers": {
    "qualito": {
      "command": "uvx",
      "args": ["qualito-mcp"]
    }
  }
}

Tools available: qualito_setup, qualito_score, qualito_cost, qualito_patterns, qualito_warnings, qualito_templates, qualito_incidents, qualito_slo.

Local Development

git clone https://github.com/mp-web3/qualito.git
cd qualito
uv sync --extra dev --extra server

# Run tests
uv run pytest

License

MIT