Why Dewey exists

The problem

AI coding assistants read your docs before they write a single line of code. If your documentation is scattered, outdated, or written only for humans skimming a README, agents will hallucinate, guess, or silently get things wrong.

Most documentation tools optimize for presentation — typography, sidebars, search. None of them ask: can an LLM actually parse this?

Agent-ready docs

Dewey treats documentation as structured data, not just rendered pages. Every doc page can have a human-readable version and a dense, machine-optimized .agent.md counterpart — tables over prose, explicit entry points, no ambiguity.

The output isn't just a website. It's AGENTS.md, llms.txt, and install.md — files that coding assistants already know how to find and use.

Preparation over presentation

Dewey is a docs agent, not a docs framework. It audits your markdown against a 100-point rubric, scores it, and tells you exactly what's missing. Then it generates the agent-facing files automatically.

The philosophy is simple: spend your energy on content quality, not on configuring a static site generator. Dewey handles the scaffolding so you can focus on what you're documenting.

Who's behind this

Dewey is built by Arach, a software engineer and engineering leader who's spent the last decade building developer tools, leading engineering teams at Meta, and co-founding startups as CTO.

After watching AI assistants struggle with real-world codebases — not because the models are bad, but because the docs are — Dewey became the obvious thing to build.

Handcrafted in Montreal, NY, and SF.