ussumant/llm-wiki-compiler

Claude Code plugin that compiles markdown knowledge files into a topic-based wiki. Implements Karpathy's LLM Knowledge Base pattern.

What it solves

LLM Wiki Compiler reduces the context window costs and fragmentation of AI agents by synthesizing scattered raw data—such as markdown files, codebases, and external bookmarks—into a structured, topic-based wiki. Instead of an agent re-reading hundreds of raw files in every session, it reads a synthesized index and specific topic articles, which can reduce token usage by up to 90%.

How it works

The tool operates as a plugin for Claude Code and Codex, implementing a "Knowledge Base" pattern. It follows a specific pipeline:

  1. Ingestion: It collects raw sources from local directories, code repositories (scanning READMEs, API contracts, and ADRs), or external services like X bookmarks.
  2. Compilation: An LLM compiles these sources into a set of topic-based markdown articles with backlinks to original sources.
  3. Consumption: The AI agent reads the compiled wiki (starting with INDEX.md) at the beginning of a session to gain a synthesized understanding of the project.
  4. Maintenance: The wiki is updated incrementally as sources change, and users can visualize the knowledge graph or "lint" the wiki for contradictions and orphans.

Who it’s for

Developers and knowledge workers using AI agents (like Claude Code or Codex) who manage large projects with extensive documentation, complex codebases, or fragmented research notes.

Highlights

  • Codebase Mode: Automatically discovers project architecture, API contracts, and decision records from a repository to create a technical wiki.
  • Context Reduction: Drastically lowers token costs by replacing raw file reads with synthesized articles.
  • Knowledge Graph: Includes an interactive, canvas-based visualization of topics and concepts.
  • External Integration: Supports fetching bookmarks from X (Twitter) and provides a framework for adding other connectors like Readwise or Slack.
  • Dual-Layer Storage: Supports both a global wiki for cross-project research and local wikis for project-specific knowledge.

Related

  • Project
    sdyckjq-lab/llm-wiki-skill

    A system for building a structured, interconnected knowledge base from fragmented data, featuring an interactive knowledge graph and integration with AI CLI agents.

  • Project
    SamurAIGPT/llm-wiki-agent

    A coding agent skill that converts source documents into a persistent, interlinked markdown wiki with automated entity extraction and contradiction flagging.

  • Project
    Astro-Han/karpathy-llm-wiki

    A reusable agent skill that enables LLMs to build and maintain a structured wiki of synthesized knowledge, moving beyond simple RAG by compiling sources into durable, cross-linked markdown pages.

  • Project
    nvk/llm-wiki

    llm‑wiki is a cross‑runtime tool that lets Claude, Codex, OpenCode, Pi and other LLM agents build and query a structured, Obsidian‑compatible wiki. It provides ingestion, parallel research, session memory, private adapters, and exportable knowledge checkpoints, with install scripts for each supported agent.

  • Project
    atomicstrata/llm-wiki-compiler

    A Node.js tool that compiles raw documents into a citation‑aware, typed Markdown wiki with lifecycle profiles, review gates, and hybrid retrieval; usable via CLI, local UI, MCP server, or TypeScript SDK.