LLM Wiki Agent is a knowledge base automation tool that builds and maintains a structured wiki from source documents using Claude, Codex, or Gemini as the underlying agent.
The tool solves the problem of knowledge fragmentation by automating wiki construction. Instead of manually organizing notes, you drop source documents into a folder and the agent reads them, extracts key entities and concepts, and builds an interlinked markdown wiki. The wiki grows richer with each new source, with cross-references and synthesis already in place. The agent detects contradictions between sources at ingest time and maintains a knowledge graph showing relationships between pages.
The tool suits researchers building references over time, readers working through books chapter by chapter, and anyone maintaining a personal knowledge base who wants structure without manual curation. It works with diverse file formats including markdown, PDF, DOCX, PPTX, XLSX, HTML, TXT, CSV, JSON, XML, RST, and EPUB, converting non-markdown files automatically at ingest. The project positions itself against traditional knowledge tools that require you to search your own notes, instead delivering a synthesized wiki that compounds over time with entity pages for people and organizations, concept pages for key ideas, a living overview that updates with each ingest, contradiction flags, and lint reports identifying orphan pages and data gaps.
The project shows active development with multiple specialized features being maintained. The codebase includes slash commands specific to Claude Code alongside natural language triggers that work across all supported agents. The tool integrates with a knowledge graph visualization that uses community detection to cluster related topics. The project maintains documentation of use cases spanning research, book reading, and personal knowledge management, suggesting ongoing refinement based on real usage patterns.
