sage-wiki

A self-hosted AI knowledge base turns documents into linked Markdown articles and a knowledge graph that people and agents can query through MCP.

sage-wiki turns a document collection into a linked wiki and knowledge graph for researchers, teams, and AI agents that need reusable knowledge across tasks.

Document compilation. Build connected articles from notes, PDFs, Office documents, EPUBs, email, code, and transcripts; image descriptions require a vision-capable model such as Gemini, Claude, or GPT-4o.
Search and graph queries. Search combines text matches, semantic similarity, and graph connections, while relational queries follow links between facts.
Evidence and history. Optional graph extraction attaches supporting passages, source documents, and confidence to relations; historical queries retrieve what the graph held at an earlier point.
Agent memory. MCP lets agents search, query, capture knowledge, and compile sources, with generated skill files guiding when to use those actions.
Readable wiki. Browse the same collection through a web interface, terminal interface, or Obsidian-compatible Markdown, with review-gated alias resolution for names such as K8s and Kubernetes.

The MIT-licensed open-source software runs as a Go binary on your own machine or server. Its default graph uses keyword proximity; evidence-bearing relations require optional model passes. Those passes add model calls, and entity merge proposals require review by default.