Open Source — MIT Licensed

The open source AI
code quality engine.

Scan, score, and improve your codebase — with zero regression risk. MIT licensed.

View on GitHub See Pricing

We scored 98/100 on our own codebase  ·  MIT licensed core  ·  5-minute setup

Overall
98
/ 100
ratchet on itself
Testing
98
/ 100
Error Handling
99
/ 100
Security
97
/ 100
Type Safety
100
/ 100
Performance
96
/ 100
Code Quality
99
/ 100
Open Source Core

The engine that scored 98/100 on itself is yours.

The full quality engine is MIT licensed and free forever. Autonomous fixing is available on paid plans — or bring your own key.

Free forever — MIT Licensed

  • ratchet scan MIT
  • ratchet vision MIT
  • ratchet report MIT
  • ratchet badge MIT
  • ratchet tighten MIT
  • ratchet status MIT
  • ratchet log MIT
  • ratchet init MIT
  • GitHub Action MIT

Autonomous fixing — Ratchet Pro

  • ratchet torque AI-powered refactoring engine ⚡ Pro
  • ratchet improve one-click quality improvement ⚡ Pro

Both commands support --local BYOK fallback — use your own OpenAI or Anthropic key without a paid plan.

How it works

Score. Fix. Ship.

Three steps to a healthier codebase — starting from zero config.

1

Scan

Add the free GitHub Action or run ratchet scan locally. Every commit gets an instant code quality score across 6 categories: Testing, Error Handling, Security, Type Safety, Performance, and Code Quality.

Free — MIT Licensed
2

Fix

Upgrade to Ratchet Pro (or use --local BYOK) and let the engine propose targeted improvements — test coverage, hardened error handling, type safety. Each click lands a real, reviewable code change.

Pro
3

Ship

Ratchet opens a pull request with the improvements, a full diff, and a before/after score card. You review, merge, and your score goes up. Repeat until you're proud of the number.

Automated
See it in action

Real improvements. Real scores.

Run ratchet improve and watch your score climb — with full diff visibility before anything merges.

Before
72
/ 100  ·  Testing: 54
Error Handling: 61
→
After
84
/ 100  ·  Testing: 78
Error Handling: 82

+12 points in one improvement cycle  ·  3 PRs auto-opened  ·  Zero regressions

Scoring system

One number. Six categories.

Ratchet's score is a weighted composite across the six dimensions that actually matter for production readiness.

ratchet score 98 / 100
Testing 25%
98
Error Handling 20%
99
Security 15%
97
Type Safety 15%
100
Performance 10%
96
Code Quality 15%
99

Embed the live badge in your README and let contributors see the score at a glance.
[![Ratchet Score](badge-url)](https://github.com/giovanni-labs/ratchet)

Pricing

Start free. Scale when ready.

The OSS engine is free forever. Upgrade for autonomous improvements and team features.

Free
$0
Unlimited scanning, forever.
  • Unlimited ratchet scan
  • Vision analysis (1/week)
  • Badge & JSON/MD export
  • GitHub Action included
  • Per-category breakdown
  • Configurable thresholds
Get Started Free
Builder
$12/mo
For developers getting started with AI fixes.
  • Everything in Free
  • 30 improvement cycles/mo
  • BYOK support
  • ratchet improve access
Get Builder
Most Popular
Pro
$35/mo
For solo devs who want the full engine.
  • Everything in Builder
  • 150 improvement cycles/mo
  • BYOK support
  • ratchet torque + improve
  • Unlimited vision analysis
  • PDF reports
Get Pro
Team
$149/mo
For engineering teams shipping together.
  • 5 seats included
  • 500 cycles/mo
  • Hosted — no API key needed
  • CI/CD GitHub Action
  • Slack & Discord notifications
Get Team
Enterprise
Custom
For large orgs with compliance needs.
  • Unlimited cycles
  • Self-hosted / air-gapped
  • Custom scoring rules
  • SLA + dedicated support
  • Audit log & SSO/SAML
Contact Us

Need more cycles? Add extra capacity at $0.75 / cycle on any paid plan.

GitHub Action

Add to any repo in 60 seconds.

Drop this into .github/workflows/ratchet.yml and you're done. Free and open source.

# .github/workflows/ratchet.yml
name: PR Quality Gate

on:
  pull_request:
    branches: [main]

permissions:
  contents: read
  pull-requests: write

jobs:
  quality:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Ratchet Code Quality Scan
        uses: giovanni-labs/ratchet@v1
        with:
          threshold: 80
          category-thresholds: Security=85,Testing=70
FAQ

Common questions

Everything you need to know about Ratchet.

Is Ratchet fully open source?
The core engine (scan, quality scoring, GitHub Action, vision, badge) is MIT licensed and free forever. Autonomous fixing (torque, improve) is available on paid plans, with a --local BYOK fallback so you can use your own OpenAI or Anthropic API key without subscribing.
Can I use Ratchet in CI/CD?
Yes. The GitHub Action for scan + tighten is free and open source. Run ratchet scan --fail-on 80 in any pipeline to enforce a quality gate on every PR. No account required.
How accurate is the 98/100 self-score?
We dogfood Ratchet on the Ratchet codebase itself. The 98/100 is a real scan result across all 6 weighted categories. Scores are deterministic for the same commit — you can verify by cloning the repo and running ratchet scan yourself.
What are the 6 scoring categories?
Testing (25%), Error Handling (20%), Code Quality (15%), Security (15%), Type Safety (15%), and Performance (10%). Each category is scored 0–100 and weighted to produce the composite score. Weights reflect real-world impact on production reliability.
What is BYOK / --local mode?
Bring Your Own Key. Pass --local to ratchet torque or ratchet improve along with your own OPENAI_API_KEY or ANTHROPIC_API_KEY and the engine runs entirely on your API quota. No Ratchet Pro subscription needed for local usage.
Does Ratchet support monorepos or multiple languages?
Yes. Ratchet analyzes any codebase structure and supports TypeScript, JavaScript, Python, Go, Rust, and more. Use ratchet init to configure paths, exclusions, and custom scoring rules per-project.