LLM Wiki Compiler is an open-source knowledge compiler CLI with 2,200+ stars on GitHub: feed it raw sources and it compiles them into an interlinked, LLM-maintained wiki, following Andrej Karpathy’s LLM Wiki pattern.
Personal knowledge bases usually rot: notes pile up unlinked, and retrieval quality depends on how disciplined you were on the day you wrote them. This tool, published on npm as llm-wiki-compiler and updated to version 1.5 in 2026, inverts the workflow — raw sources go in, and the compiler produces a structured wiki of concept pages with links between them, which an LLM can then answer questions over with citations. Recent releases separate wiki behavior from storage so existing wikis keep their layout without migration, add a review step where generated answers and their citations can be approved in batches before becoming pages, and ship a local viewer that refreshes after each compile. It integrates with the agent ecosystem too: the project speaks MCP, works alongside tools like Obsidian, and supports bring-your-own provider endpoints, making compiled knowledge available to AI agents as well as human readers.
LLM Wiki Compiler is free and open source under the MIT license. It runs locally as a CLI; you only cover your own LLM provider usage for compilation and question answering.
Researchers, writers, Obsidian users and agent builders who want their scattered sources turned into a maintained, citable knowledge base instead of another folder of forgotten notes.