Obsidian Semantic MCP
Turns an Obsidian vault into a queryable brain for Claude — semantic search across every note using local embeddings and pgvector, no cloud services required.
Obsidian Semantic MCP solves the reset problem: AI assistants forget everything between sessions. Your notes, decisions, and accumulated context sit in Obsidian but never make it into the conversation automatically. This tool closes that gap.
It indexes every note as a vector embedding via Ollama and nomic-embed-text, stores them in PostgreSQL with pgvector for fast semantic search, watches the vault for changes and re-indexes automatically, and exposes the whole thing through MCP so Claude can retrieve and write vault content mid-session without any manual copy-paste.
Everything runs locally. No API keys, no cloud round-trip, no data leaving the machine.
What it does
- Semantic search across the full vault — finds notes by meaning, not just keywords
- Full CRUD via MCP tools: read, write, search, list — works even when Obsidian is closed
- Auto-reindex on vault changes via a file watcher
- PostgreSQL + pgvector backend: persistent, queryable, fast
- One-command setup wizard (
osm init) with dry-run mode
Status
Cast. Running in the workshop as the long-term memory layer for Claude Desktop sessions.