superlore
The agent-native knowledge base. One corpus. Humans and agents.
superlore is an agent-native knowledge base. You author your knowledge once, in MDX, and the same structured content becomes two things at once:
- a clean, interactive, visual knowledge base for humans, and
- a first-class MCP (Model Context Protocol) server for agents, over the same content.
Most docs tools render pages for people and bolt an MCP onto scraped HTML afterwards. superlore is content-first: a structured content model is the source of truth, and both the human site and the agent MCP are projections of it. Nothing is scraped; the two faces never drift.
The one idea to take away
Every superlore component has two faces from one authored instance — a render face for humans
and a knowledge face (typed, structured data) for agents. A Timeline is never a picture an
agent has to interpret; the agent gets the items behind it.
Why superlore
For readers
Calm, editorial, fast. Search-first, light and dark as equals, structural components that make complex knowledge legible at a glance.
For agents
A built-in MCP exposes search, pages, sections, lists, relations, and the data behind every component — no scraping, no guessing.
For authors
Write MDX. Drop in cards, timelines, boards, entities, tables, and diagrams. Deploy anywhere.
Start here
Get started
Let Claude build a KB, or scaffold one from the terminal.
Architecture
How content-first + the dual-representation contract fit together.
Authoring
What a superlore document is, and how to write one.
Authoring for agents
How an agent should write superlore-style docs.
What you get
This very site is built with superlore and is MCP-enabled and public — point your agent at
/api/mcp and it can read these docs directly.