Axiom Wiki

Axiom Wiki

AI-powered personal knowledge base that automatically ingests documents and maintains an interconnected wiki of markdown pages. Inspired by Andrej Karpathy's llm-wiki.

2024 - Present

Available on NPM

Open Source

Overview

Axiom Wiki is a high-performance personal knowledge management system that shifts away from traditional RAG (Retrieval-Augmented Generation) by compiling knowledge into a persistent, compounding wiki.

Unlike standard RAG systems that re-derive answers from raw sources for every query, Axiom Wiki extracts entities, concepts, and cross-links to build a structured markdown-based knowledge graph. It supports diverse input formats including PDFs, images, DOCX, and even entire codebases.

Designed with developer productivity in mind, it features MCP (Model Context Protocol) integration, allowing it to serve as a high-fidelity context provider for AI agents like Claude Code or Cursor.

Key Features

Automated Ingestion

Extracts entities, concepts, and cross-links from PDFs, markdown, images, DOCX, and HTML files automatically.

Codebase Mapping

Scans entire projects to generate structured documentation, including architecture overviews and module relationships.

Incremental Sync

Uses git diff to detect changes and re-generates only the affected wiki pages, keeping documentation perfectly in sync.

Web Clipper

Integrated CLI command to fetch and clean web articles via Readability for instant knowledge ingestion.

MCP Mesh Support

Acts as a Model Context Protocol server, enabling AI agents to query your personal wiki as a tool.

Obsidian Compatible

Generates plain markdown with frontmatter, making it perfectly suited for Obsidian vaults and local-first workflows.

Technology Stack

Core & Runtime

TypeScript Node.js (>= 18) pnpm Commander.js

AI Providers

Google Gemini OpenAI Anthropic Ollama (Local)

Tooling & Infrastructure

Vitest Starlight (Astro) Docker GitHub Actions