Skip to content

Metrics Overview

Every metric in AST Metrics exists to answer one practical question about your code. You don't need to learn them all: start from the question you're asking, and read the page that answers it.

If you have never run an analysis, do the hands-on tutorial first: it walks through a real analysis of Monolog and teaches you to read the numbers in ten minutes. Every page below then goes deeper on one metric, always in the same shape: what it measures, how to read it, and what to do about it.

"Where should I start refactoring?"

  • Risk Score crosses complexity with git activity and points at the handful of files where the next bug is statistically waiting. This is the first list to read on any codebase.

"Is this code too complicated?"

  • Cyclomatic Complexity counts the independent paths through a function: every path is something to understand and to test.
  • Volume counts lines of code, logical lines and comments. Basic, but size is the metric that correlates most strongly with defects.
  • Maintainability Index blends both into a single 0-100 health score per file, perfect for trends and thresholds.

"Is this class well designed?"

  • Class Cohesion (LCOM4) detects classes that are really two or three unrelated classes stuck together.
  • Coupling & Instability measures who depends on whom, and whether your dependencies point in a direction that keeps change cheap.

"Do my tests actually protect me?"

  • Test Quality measures test isolation, god tests, and the production classes that have no test at all, without running a single test.

"Does my architecture match what I think it is?"

  • Architecture Map reads your real structure and states it in plain words: the communities your classes form, the layered map, the cycles with the exact dependencies to cut, and the rules that freeze the boundaries in CI.
  • Community Detection is the deep dive behind it: how the communities are found, every finding, the git history read against the boundaries.

"What happens if someone leaves?"

  • Bus Factor reads the git history to find the parts of the codebase only one person knows.

Every metric is available in the terminal summary (ast-metrics analyze), the HTML report (--report-html), the JSON report, and to AI agents through the MCP server. Most of them can also be enforced as lint rules with thresholds you choose.