Obsidian Semantic Search
Let your coding agent semantically search your Obsidian vault locally so specs, notes, and research surface during idea work and implementation without uploading notes to a cloud RAG service.
Overview
Obsidian Semantic MCP is a MCP server for the Idea phase that provides read-only semantic search over a local Obsidian vault using Ollama embeddings.
What is this MCP server?
- Read-only semantic search over a local Obsidian vault via MCP
- Embeddings through local Ollama (default model bge-m3)
- External SQLite index under configurable OBSIDIAN_SEMANTIC_MCP_HOME
- Required OBSIDIAN_VAULT_ROOT absolute path to the vault
- Optional startup incremental indexing via OBSIDIAN_SEMANTIC_STARTUP_INDEX
- Server version 0.2.1
- Default embedding model bge-m3 via OBSIDIAN_EMBED_MODEL
- Read-only access to vault; index stored outside vault when OBSIDIAN_SEMANTIC_MCP_HOME is set
What problem does it solve?
Your best product context lives in Obsidian, but coding agents cannot search it by meaning unless you paste notes or ship data to a hosted search API.
Who is it for?
Builders with an existing Obsidian PKM who run Ollama locally and want private, read-only semantic retrieval in the agent loop.
Skip if: Teams without Obsidian, without a local Ollama setup, or who need agents to edit vault files rather than search them.
What do I get? / Deliverables
After configuration, your agent queries the vault through MCP and gets semantically related note chunks from a local SQLite index backed by Ollama.
- Queryable semantic index of vault markdown in external SQLite storage
- MCP tools that return relevant note content from natural-language queries
- Optional incremental re-index on server startup
Recommended MCP Servers
Journey fit
PKM-backed research is where solo builders capture competitors, audience notes, and product thinking before and during build. Semantic vault search is a read-only research affordance: find related notes by meaning, not filenames, while scoping what to build.
How it compares
Local PKM semantic search MCP, not agent memory storage or cloud analytics.
Common Questions / FAQ
Who is Obsidian Semantic MCP for?
Solo builders and indie hackers who document research in Obsidian and use Claude Code or Cursor with a local Ollama instance for embeddings.
When should I use Obsidian Semantic MCP?
Use it when researching ideas, scoping features from past notes, or drafting docs while you need the agent to surface related vault content without cloud upload.
How do I add Obsidian Semantic MCP to my agent?
Install @dalecb/obsidian-semantic-mcp, set OBSIDIAN_VAULT_ROOT to your vault path, ensure Ollama is reachable at OLLAMA_BASE_URL, then register the stdio server in your MCP client config.