When a data engineer spends evenings scrolling through an Obsidian vault, the notes feel like a personal encyclopedia. The pages are searchable with a quick Ctrl+F, but the moment you want to feed that knowledge into an LLM or a data pipeline, the workflow stalls. You end up copying snippets, maintaining separate markdown files, or writing custom parsers that never quite capture the context you need. The gap between a rich, locally stored knowledge base and a programmatic query surface is the hidden cost of a read-only dependency.
Turning a Static Vault into a Living Knowledge Store
LoreDocs removes that friction with a one-time import that turns any Obsidian vault into a fully queryable repository. Point the vault_import_dir at the root of your Obsidian directory, and the import routine walks every subfolder, extracts YAML front-matter tags, and creates a single SQLite file that contains every note, its metadata, and a complete version history. Because the data lives in a local SQLite file, you retain full ownership—no cloud lock-in, no external service required.
The import process respects the natural organization of your vault. Each top-level folder can become a named vault, and you can assign tags that act as lightweight classifiers across vaults. The result is a multi-vault knowledge store where each vault can be gated by tier, allowing you to keep experimental drafts separate from production-grade documentation. Since the SQLite file is portable, you can move the entire knowledge base to a new laptop, a CI runner, or a secure server with a single copy operation.
Document versioning is baked in. Every edit creates a new revision, and the restore capability lets you roll back to any prior state without leaving the LoreDocs interface. This versioned approach mirrors the way data pipelines handle schema migrations, giving you a reliable safety net when knowledge evolves.
Search That Grows With Your Projects
Once the vault is inside LoreDocs, the search experience expands far beyond a simple keyword match. The built-in FTS5 full-text engine indexes every word across all vaults, so a query like "latent drift detection" instantly surfaces every paragraph that mentions the phrase, regardless of which vault it lives in. The vault_prime primitive lets you inject the entire context of a selected vault into a single API call, meaning an LLM can answer questions with the full background it needs, not just a handful of snippets.
For teams that need deeper semantic understanding, the Pro tier adds a hybrid search powered by LanceDB. The engine combines BGE embeddings with BM25, then fuses the results using reciprocal rank fusion. Documents are split at paragraph boundaries into chunks of up to 256 tokens, preserving natural context while keeping the index efficient. A call to vault_search with semantic=true returns results that rank not only by keyword relevance but also by conceptual similarity, a capability that turns a static note collection into an active knowledge assistant.
The workspace-scoped auto-vault feature further reduces onboarding friction. By calling vault_open_workspace(path), the system either creates a new vault bound to the given directory or returns the existing one on subsequent calls. This mirrors the familiar workspace model many data engineers already use, so a new project can start with a ready-made knowledge store without manual configuration.
Seamless Integration With Claude and Other Agents
One of the most compelling reasons to adopt LoreDocs is its cross-vendor MCP compatibility. The product ships as a native MCP server that works out of the box with Claude Code, the OpenAI Codex desktop app, and the Cursor IDE. All you need is a project-local .mcp.json (or .cursor/mcp.json for Cursor) placed in the root of your codebase. The MCP definition tells each client how to reach the LoreDocs server, and the verification logs confirm that the integration works without any per-client tweaks.
When an environment cannot read the MCP file, the Python fallback script query_loredocs.py provides a pragmatic bridge. Any agent that can execute Python can import the script, point it at the SQLite file, and issue the same search and retrieval commands that the MCP interface offers. This dual path ensures that whether you are building a custom data pipeline, a Jupyter notebook, or an autonomous agent, the knowledge base remains reachable.
The CLI, built with Click, gives you direct control from the terminal. Commands like vault list, doc add, search, and archive let you manage vaults and documents without leaving the shell. The 43 MCP tools bundled with LoreDocs cover the full lifecycle—from onboarding a new workspace to rebuilding a semantic index and linking session output back to source documents. For teams that automate their workflows, these tools become building blocks for reproducible pipelines that treat knowledge as a first-class data asset.
Controlling Cost While Keeping Full Control
LoreDocs offers a free tier that includes three vaults, enough for most personal projects or small teams experimenting with LLM-augmented workflows. When you need unlimited vaults, the Pro tier is priced at nine dollars per month, a modest fee that unlocks semantic search, auto-discovered document relationships, and the full suite of MCP tools. Because all data lives locally, there are no hidden storage fees, and you can run the server on any machine that can host SQLite.
The local-first design also satisfies security and compliance requirements common in data engineering environments. Since the SQLite file is owned by the user, you can encrypt it, back it up to an internal repository, or place it behind your organization's firewall. There is no need to grant external services access to proprietary documentation, model prompts, or internal data schemas.
Bring Your Knowledge Into the Loop
If you have spent years curating an Obsidian vault and are ready to let that knowledge power your LLM workflows, LoreDocs provides a clear path from static markdown to an interactive, searchable store. The import is a single command, the search primitives are ready to call from Claude or any Python script, and the optional Pro features give you semantic depth when you need it. Start with the free tier to explore the basics, and upgrade when your projects demand richer indexing and full MCP integration.
Ready to make your vault queryable? Visit the tools page to see LoreDocs in action, or reach out through the contact form to discuss a tailored onboarding plan.