rabit

Burrows for the Agentic Web

A simple convention for publishing content that humans can read and AI agents can navigate reliably. Like Gopher, but for the modern AI era.

{
  "version": "1.0",
  "title": "My Documentation Burrow",
  "entries": [
    { "type": "link", "title": "Getting Started", "uri": "/start.md" },
    { "type": "burrow", "title": "API Reference", "uri": "/api/.burrow.json" }
  ]
}
A simple .burrow.json manifest

Core Concepts

Structure your content for reliable discovery

🕳️

Burrows

A collection of content organized via a `.burrow.json` manifest. Think of it as a directory listing that agents are guaranteed to understand.

🐰

Warrens

A registry or collection of Burrows. A way to index multiple documentation sites or repositories in a central place.

🧭

Navigation

Agents follow links in manifests rather than guessing URLs or scraping HTML soup. Deterministic traversal.

📄

Transport Agnostic

Works over HTTP, Git, or local files. The protocol is the JSON schema, not the transport layer.

Why Rabit?

HTML Web

Messy, Visual, Unstructured

vs

Rabit Web

Structured, JSON-first, Reliable

The modern web is built for human eyes. AI agents struggle with complex layouts, dynamic JavaScript, and inconsistent navigation. Rabit provides a parallel structure—a "burrow"—underneath your site that agents can traverse efficiently.

Use Cases

📚

Documentation

Provide a perfect navigation map for AI coding assistants to read your docs.

🧠

Knowledge Bases

Structure internal wikis so they can be easily ingested by RAG systems.