Is this the same as customer support knowledge base software?

No. Help-desk knowledge bases optimize for deflecting repeat questions with the right existing answer. A research knowledge base has the opposite objective — surfacing what the material has not yet said — and the tooling differs all the way down, from how content is structured to how retrieval is scored.

Can I keep my AI knowledge base local and open source?

Yes. The base is your own folder of markdown files and PDFs, optionally under Git version control, in formats that stay readable without any particular application. The InfraNodus skills that build the semantic layer are open source, and the ontology files they write are plain markdown.

What is a content gap in a knowledge base?

A content gap is a pair of topical clusters that both exist in your knowledge base but are not connected to each other. Gaps are detected with community detection and betweenness centrality rather than by asking the model, so the result is reproducible, and a question that bridges two clusters tends to produce something the material implies but never states.

AI Knowledge Base

Q: What is an AI knowledge base?

An AI knowledge base is your own material — notes, papers, transcripts, datasets — stored in a form a large language model can read, query, and extend. In practice it has three layers: files in folders, a semantic layer describing the concepts and their relations, and an AI agent pointed at both.

Q: Do I need a knowledge graph, or is retrieval enough?

If your knowledge base is reference material and your questions are lookup, vector retrieval is enough and additional layers will slow you down. If you are doing research, a knowledge graph provides multi-hop reasoning across related concepts and gap detection between topics, neither of which similarity search can produce.

An AI knowledge base can function as a second brain for your personal or professional use. The basic setup is a folder of files, a semantic layer describing the concepts and their relations, and an AI agent pointed at both.

In this section we will cover the basic setup and the tools you need to create and use an AI knowledge base.

1. Get InfraNodus MCP 2. Install the InfraNodus Skill
3. VSCode/ Cursor Graph View Plugin 3. Obsidian Graph View 4. KB Installation Guide

Please, choose your preferred setup:

Personal Knowledge Base

You have nothing yet. Files, folders, markdown, version control, and a viewer. No programming.

AI Second Brain

You want the whole loop. Sources in, structure read, gaps surfaced, questions out.

LLM Wiki

You are working one corpus hard. Karpathy's pattern, with the graph layer wired in.

AI Reasoning Graph

You need the model to reason, not retrieve. Reasoning ontologies in plain text — types, constraints, perspectives.

RAG vs. LLM Wiki vs. Knowledge Graph

Your AI only summarizes your vault. Choosing the semantic layer that fixes it.

The Three Layers: Storage, Semantic Layer, and AI Query Interface

Every working setup has the same three layers, usually built in this order.

First, you need to organize your files and documents using the Zettelkasten method, in traditional folders, or both. If you prefer not to have any hierarchy, you can also use tags and [[backlinks]] (Roam Research style).

The second layer is the semantic layer, which is a machine-readable description of the concepts and how they relate: an LLM wiki, a knowledge graph, or both. Without one, the model falls back on vector similarity. Add typed relations, constraints, and perspectives and it becomes a reasoning ontology — structure that shapes how the model thinks, not only what it finds.

The third layer is the AI query interface, which is Claude Code, Codex, or the AI in InfraNodus, pointed at the folder with that layer available as a tool.

An AI knowledge base represented as a knowledge graph of concepts, topics, and the gaps between them — A knowledge base and the graph of what is inside it — topics in colour, gaps between them.

Layer	What it is	Covered in
Storage	Files in folders. Markdown and PDFs, a viewer — Obsidian, Cursor, Notion, or InfraNodus — and Git if you want version history.	Personal knowledge base
Semantic layer	A machine-readable description of the concepts and how they relate: an LLM wiki, a knowledge graph, or both. Without one, the model falls back on vector similarity. Add typed relations, constraints, and perspectives and it becomes a reasoning ontology — structure that shapes how the model thinks, not only what it finds.	LLM wiki · Reasoning graph · Choosing one
Query	Claude Code, Codex, or the AI in InfraNodus, pointed at the folder with that layer available as a tool.	AI second brain

What the Graph Adds

Storage, retrieval, and a wiki all describe what a knowledge base contains. None of them describes its shape. When represented as a network, the knowledge base becomes measurable: the graph helps find relations between remote concepts, betweenness centrality finds the concepts holding the material together, community detection finds the topics that have actually formed, and pairs of topics that both exist without being connected are the content gaps — relations your material implies and never states.

A knowledge base that only accumulates converges. What makes it generative is knowing where it is not yet connected.

Tools

Install the InfraNodus skill, connect the MCP server to Claude, open your folder, and run /infranodus on one subfolder rather than the whole thing. You get a graph, a manifest.json your AI reads afterwards, and a report with the topics, the gaps, and the questions worth asking.

InfraNodus skill — graphs for the folders of an existing vault.
LLM Wiki skill — the full wiki scaffold, with gap detection at each stage.
Ontology Creator skill — a reasoning ontology in wikilink syntax, from a prompt or a corpus.
MCP server — network analysis and GraphRAG as tools for your LLM.
Obsidian plugin graph view plugin for a popular markdown editor Obsidian.
VSCode / Cursor AI extension — graph view plugin for VSCode and Cursor — popular IDEs that can be more useful than Obsidian for maintaining AI-powered knowledge bases.
API and GraphRAG endpoints — for your own pipeline.

FAQ

What is an AI knowledge base?

Your own material — notes, papers, transcripts, datasets — stored so a large language model can read, query, and extend it. In practice: files in folders, a semantic layer describing the concepts and their relations, and an AI agent pointed at both.

How do I create a knowledge base for AI?

Put your sources in folders, convert what you can into markdown, and point Claude Code at the folder. Then add a semantic layer so the model has a structural view rather than the nearest matching paragraph. Details in the setup guide.

Is this customer support knowledge base software?

No. Help-desk knowledge bases optimize for deflecting repeat questions with the right existing answer. This is the opposite objective, and the tooling differs all the way down.

Do I need a knowledge graph, or is retrieval enough?

If your base is reference material and your questions are lookup, retrieval is enough. If you are doing research, the graph is what gives you multi-hop reasoning and gap detection — see the comparison.

Can I keep it local and open source?

Yes. Markdown files and PDFs in a folder you own, optionally under Git. The InfraNodus skills are open source and the ontology files they write are markdown.

What is a content gap?

A content gap is identified between two topical clusters that both exist in your knowledge base but are not connected to each other. Detected with community detection and betweenness centrality, so the result is reproducible rather than a model's guess.

Build Your AI Knowledge Base

Connect your notes, papers, and files to a knowledge graph that shows you the topics inside them, the concepts that hold them together, and the gaps between them: