Nouz
Connect an Obsidian vault to MCP so agents can query a semantic knowledge graph with auto-classification and DAG hierarchy instead of grepping thousands of notes.
Overview
NOUZ-MCP is an MCP server for the Build phase that exposes a semantic, auto-classified knowledge graph over an Obsidian vault to AI coding agents.
What is this MCP server?
- Semantic knowledge engine for Obsidian with auto-classification and DAG hierarchy
- Stdio MCP server nouz-mcp v3.1.0 from PyPI
- OBSIDIAN_ROOT vault path with SQLite cache (default obsidian_kb.db)
- Configurable embeddings via EMBED_ENABLED, EMBED_PROVIDER, and related env vars
- Optional NOUZ_CONFIG yaml for explicit server configuration
- Server version 3.1.0
- PyPI package identifier nouz-mcp
- Default SQLite cache filename obsidian_kb.db
Community signal: 15 GitHub stars.
What problem does it solve?
Your Obsidian graph is rich for humans but too large and unstructured for agents to query reliably with simple search.
Who is it for?
Indie builders who already use Obsidian as a product and research hub and want DAG-aware semantic retrieval in Claude Code or Cursor.
Skip if: Builders without Obsidian, or anyone who only needs one-off PDF or web page ingestion with no note graph.
What do I get? / Deliverables
After install, agents query classified, hierarchy-aware knowledge from your vault via MCP with a local SQLite cache for faster repeats.
- MCP tools backed by Obsidian semantic index and DAG hierarchy
- Local SQLite knowledge cache under the vault
- Embedding-enhanced classification and query when configured
Recommended MCP Servers
Journey fit
Build is canonical because you register NOUZ while wiring agent tooling to your second-brain vault during product and research work. Agent-tooling fits the Obsidian MCP bridge, SQLite cache, and embedding-powered classification that extend what coding agents can retrieve.
How it compares
Obsidian semantic graph MCP, not LINZA’s generic Markdown vault sidecar nor a stock filesystem list_dir integration.
Common Questions / FAQ
Who is NOUZ-MCP for?
Solo builders and researchers who organize work in Obsidian and want MCP-backed semantic search and classification for coding agents.
When should I use NOUZ-MCP?
Use it when your agent needs structured answers from a large vault—specs, competitors, audience notes—without manually opening dozens of files.
How do I add NOUZ-MCP to my agent?
Set OBSIDIAN_ROOT (and optional NOUZ_CONFIG or database paths), configure embeddings if needed, run nouz-mcp via PyPI/stdio, and register it in your MCP client config.