Kura Docs
Get started / Introduction

Introduction

Kura is the knowledgebase for humans and agents, built on June. One source of content: humans read a polished website, while AI agents query and call it in a structured, machine-readable way — single source, never out of sync.

Why Kura

The audience for docs is changing: more and more traffic comes from AI agents. Traditional docs tools treat "for agents" as a bolt-on; Kura makes it the foundation.

One source → HTML (for people), Markdown and JSON (for machines), MCP tools (for agents to call), with zero duplication.

MDX components

This tip box is written with the MDX <Callout> component — people see a styled card, while the agent-facing .md is clean plain text.

A shared knowledgebase

  • Humans: sidebar, table of contents, search, copy-as-Markdown browsing.
  • Agents: every page has a .md source, a /llms.txt index, and a callable /mcp search tool.

Design principles

  1. Deploys anywhere out of the box — a filesystem-free Worker bundle with zero-dependency lexical search; no model or service to provision.
  2. Semantic search when you want it — add an embedder (bge-m3, multilingual) and the same pure-JavaScript retrieval engine upgrades from keywords to meaning.
  3. Convention over configuration — what Mintlify makes you bolt on, Kura ships built in.