Skip to content
unfault

Welcome to Unfault

Unfault is an open-source CLI that analyzes your code for patterns that cause production problems.

It’s not a linter. It doesn’t police style or enforce conventions. It looks at how your code behaves at runtime: what calls can hang, where errors disappear, what fails under load, what breaks when a dependency goes down.

Run it before you push. Run it in CI. Pipe it into your AI agent. It tells you what’s actually there.

Commands

Section titled “Commands”
  • unfault review: narrative summary with the most important findings, each named and explained
  • unfault lint: all findings grouped by severity and rule, linter-style
  • unfault graph: impact analysis, dependency queries, hub file detection
  • unfault config: configure LLM providers and inspect observability integrations
Section titled “Quick Links”
Installation Get unfault running in under a minute.
Quick Start Run your first review and understand the output.
Rules Catalog Browse all the patterns unfault detects.
How It Works What gets parsed, what gets analyzed, what stays on your machine.

Use It Where You Work

Section titled “Use It Where You Work”
  • Terminal: unfault review before you push. unfault lint to see everything.
  • CI/CD: Gate on findings with exit codes. SARIF output for GitHub Code Scanning.
  • AI Agents: --output json and graph commands are built for agent consumption. See Use with AI Agents.

Philosophy

Section titled “Philosophy”

Production failures are usually visible in the code before they happen. The HTTP call with no timeout. The retry loop with no backoff. The error handler that swallows the exception. These aren’t hard to spot. They just don’t get spotted because nobody’s looking for them systematically.

Unfault looks for them. It won’t catch everything, and it doesn’t try to. It focuses on the patterns that have actually caused production incidents, explains the tradeoff in each case, and lets you decide what to do about it.