Knowledge graph from documents with ontology

general

mykg is a Claude Code skill that automatically builds a confidence-scored knowledge graph from mixed document collections — Markdown, PDF, Word, PowerPoint, Excel, HTML, images, and plain text — grounded in an induced RDFS/OWL ontology. The extraction pipeline supports LLM providers Anthropic, OpenAI, Ollama, and OpenRouter, and every stage persists intermediate state so runs can be resumed or restarted from any step after a crash or edit. Outputs include an Obsidian Vault of wikilinked notes, Neo4j LOAD CSV files, and an MCP server exposing 13 query tools for entity search, relationship exploration, and graph traversal in Claude Desktop or any MCP-compatible client. Sessions can be merged or extended incrementally — only changed files are reprocessed. The tool is aimed at developers and data engineers who need to turn unstructured corporate or research documents into a queryable, ontology-backed knowledge graph.