Introduction

Kontrua Documentation

Kontrua verifies that your repository's context is still true for AI coding agents. It reads the files you already wrote - CLAUDE.md, AGENTS.md, GEMINI.md, CONTRIBUTING.md, README.md, .cursorrules and the documents it scans - checks supported commands and paths they state against the code at that commit, and publishes the result as a kontrua/context commit status you can require in branch protection. No model takes part in that verdict.

The verdict is free and unlimited. Correcting what it found is the metered part, and nothing is ever pushed silently: on the default trigger the fix is committed into the pull request that caused it, so you review it alongside the code change that broke it.

Human docs are audited too, not rewritten. The README is verified every time the verdict runs and corrected line by line when it drifts; the architecture doc and the changelog are generated when you ask for them. API reference generation is retired: an existing API document is audited like any other file and never written.

You can also run the correction lane locally via the CLI with your own Anthropic API key - no account needed.