Build a Self-Organizing AI Knowledge Base in Obsidian

Builds and maintains a persistent, compounding knowledge base inside Obsidian so your sources, entities, and insights stay connected and cross-referenced as they grow. Unlike chat-based note assistants, claude-obsidian is a knowledge engine that creates, organizes, and evolves your wiki autonomously. Every source gets extracted and filed; every question synthesizes from the full vault; every session's context persists in a hot cache.

1. Set up an Obsidian vault with scaffold command, choosing your methodology (LYT, PARA, Zettelkasten, or generic). 2. Ingest sources (articles, PDFs, transcripts): Claude reads, extracts entities and concepts, files them into the vault, and updates the index and hot cache. 3. Ask questions: Claude reads the hot cache first, then index, then relevant pages, and synthesizes answers with citations. 4. Maintain: Run health checks to find orphans, dead links, contradictions, and stale claims. The wiki stays healthy without manual cleanup. 5. Grow: Autonomous research loops, visual canvas, and 10-principle thinking framework accelerate knowledge expansion.

Knowledge compounds like interest. Every ingest makes the vault richer; every query learns from everything that came before; every session picks up where the last one left off via hot cache, so no recap is needed.

Requires Obsidian v1.9.10+ for native Bases dashboard support (v1.6+ works with Dataview fallback). Python 3.10+ needed for retrieval pipeline and test suite. Optional: ollama (local rerank), Anthropic API key (contextual prefix tier opt-in), Local REST API plugin (MCP transport). Transport auto-detection supports Obsidian CLI, REST API, or filesystem write methods.

Web egress hygiene: /autoresearch applies URL validation and content sanitization before fetching external sources. Autoresearch can incur significant API costs (up to 45 WebFetch calls per run); budget accordingly. Methodology mode migration is manual; switching modes does not auto-migrate existing pages.