Introduction

Velu is a documentation platform built for both the people who read your docs and the AI agents that query them. You write pages in Markdown, describe your site in one velu.json file, and Velu renders a fast, themeable site with search, a table of contents, dark mode, and a built-in AI assistant.

Then it does something a static site generator can't: it helps keep those docs current. Connect a GitHub repository and Velu's AI workflows draft updates, generate changelogs, and open pull requests for you to review — so your documentation stays in sync with your product instead of drifting out of date.

Start here

How a Velu site is built

Every site is a folder with a velu.json config at its root and a set of .mdx content files:

my-docs/
├── velu.json          # Site config: name, theme, navigation
├── index.mdx          # Your home page
├── quickstart.mdx
└── guides/
    └── example.mdx

velu.json defines the theme and the navigation tree. Each entry in the navigation points to an .mdx file by its path, without the extension. When you run the dev server or publish, Velu compiles every page and wires up the sidebar, tabs, search index, and table of contents for you.

Built for AI agents, too

Every Velu site publishes its content in formats AI tools can consume: a Markdown twin of each page, an llms.txt index, an auto-generated skill.md that keeps AI coding agents aligned with your docs, and a Model Context Protocol endpoint so assistants can query your docs directly. Readers get a built-in Ask AI assistant that answers in natural language with citations to the exact page. See AI & optimization.