AFDocs v0.22.1: Test Documentation for AI Agent Compatibility

What changed

The agent-ecosystem has released version 0.22.1 of AFDocs, a tool designed to assess how well documentation sites can be understood and utilized by AI coding agents. This release implements specification v0.6.0, dated September 13, 2026. AFDocs performs checks across seven categories, evaluating aspects like content discoverability, markdown availability, and potential page size issues that could hinder AI consumption. The tool generates a scorecard, offering an overall score and detailed breakdowns for each category, alongside specific interaction diagnostics and check results. For instance, it can identify if markdown files are present but undiscoverable by agents or if large files need splitting. The project is in early development, indicated by its 0.x versioning, meaning check IDs, CLI flags, and output formats may change.

Why it matters for builders

As AI agents become more integrated into development workflows, their ability to accurately parse and act upon documentation is paramount. AFDocs offers a direct mechanism for developers to proactively ensure their documentation meets the standards required for effective AI agent interaction. This can lead to smoother onboarding for AI tools and more reliable agent performance when referencing project documentation.

Practical impact

Developers can install AFDocs globally using npm install -g afdocs or as a dev dependency for CI pipelines with npm install -D afdocs. The tool requires Node.js 22 or later. A quick start example demonstrates its usage: npx afdocs check https://docs.example.com --format scorecard. Builders can consult the full documentation at afdocs.dev for details on understanding their score, improving it with prioritized fixes, a reference for all 28 checks, CLI options, CI integration helpers, and a programmatic TypeScript API. The tool also incorporates responsible use practices, including request delays and connection caps.

Caveats and source limits

AFDocs is currently in early development (0.x), meaning its API and output formats are subject to change. The provided source does not include independent benchmark results comparing AFDocs' scores against agent performance, nor does it specify pricing information as it is an open-source tool. The latest release date mentioned in the excerpt is speculative and not tied to a real-world release.