AI writes code faster than you can review it.
Catch what got worse, in seconds.

AST Metrics measures complexity, coupling, architecture, bus factor and test quality for PHP, Go, Python, Rust, Java, C# and TypeScript. On a pull request, ast-metrics review flags only what got worse. No server, no noise, same code, same verdict.

  • Open source, MIT
  • Nothing leaves your machine

Install with brew npx pipx curl composer docker go

The whole tool is a single self-contained binary.

brew install ast-metrics/tap/ast-metrics
  • No runtime, no dependencies
  • Analysis runs locally
All installation methods (Linux packages, Windows, binaries) Quickstart guide

One command

Run it once.
Read your project like a map.

ast-metrics analyze --report-html=report . builds this report: a plain-language verdict, then every metric behind it.

Overview page of the AST Metrics HTML report: a plain-language verdict followed by scores for complexity, focused classes, method size, maintainability, test isolation and bus factor

Want to see it on real code first? Try it online, nothing to install.

What it measures.

Architecture map

Communities, layers and cycles, detected from real dependencies. Not the architecture you drew, the one you have.

Bus factor

Find the files only one person can touch, before that person leaves.

Risk hotspots

Complexity crossed with change frequency. The few files where bugs actually live.

Coupling

Afferent, efferent, instability. Keep packages easy to change in isolation.

Test quality

Isolation, traceability, god tests and orphan classes. Tests that actually protect you.

Rules and gates

Declare limits in .ast-metrics.yaml. Baseline the past, block only regressions.

Continuous integration

One line of YAML,
zero noise in your reviews.

On every pull request, the action runs ast-metrics review and comments with only the new or worsened findings. Legacy code stays quiet, and fail-on decides what blocks the merge.

GitHub Actions

PR comments, annotations, SARIF and HTML artifacts included.

GitLab CI

OpenMetrics reports and merge request gating in a few lines.

MCP server

Your metrics, readable by AI agents.

The built-in MCP server lets Claude, Cursor and other coding agents query complexity, coupling and architecture directly. Start it with ast-metrics mcp and you are done.

analyze_project find_risky_code get_coupling get_communities get_test_quality list_components + 3 more tools
Set up the MCP server

Curious what these metrics actually mean?

Every metric has its own page: what it measures, how to read it, and how to improve it.

Explore the metrics guide