# Three vaults, no account > Source: > Published: 2026-09-28 00:00:00+00:00 Data engineers and AI practitioners spend a lot of time curating prompts, model outputs, experiment logs, and internal documentation. The value of that knowledge grows when it is searchable, versioned, and easy to bring into a new project. LoreDocs is a local-first knowledge vault that keeps that data on your own disk, without requiring you to create an account. The free tier gives you three independent vaults, full-text search, and document versioning -- all stored in a SQLite file you own. The following walkthrough explains exactly what you get for free, where the limits are, and when the Pro tier becomes a practical upgrade. ### Why local and account-free matters for data-centric work When you spin up a new pipeline or experiment, the last thing you want is a cloud service that requires a separate login, API key, or ongoing subscription. A local solution fits naturally into the way you already manage code and data: everything lives in a directory that can be version-controlled, backed up, or moved between machines. LoreDocs stores metadata and search indexes in a single SQLite database while keeping the raw documents in a folder you control. You can copy the entire LoreDocs data directory to a new workstation, a container, or an external drive and pick up right where you left off -- no hidden servers, no remote sync, and no surprise charges. For a data engineer, the ability to query across multiple knowledge stores without leaving the command line or the IDE you already use can shave minutes off every debugging session. For an AI practitioner, having a searchable history of prompt experiments and model responses means you can reuse successful patterns without reinventing the wheel. The free tier is designed to give you those core capabilities without friction. ### What the free tier actually gives you Three named vaults with tag support and full-text search are the anchor. Each vault is a logical container you can label with tags, making it easy to separate project-specific notes from broader research. A keyword search for "transformer" or "spark job" returns matching documents across every vault you own instantly, because the FTS5 index lives next to the data. Document versioning is included at no cost. Every time you add or edit a document, LoreDocs records a new version so you can view change history and roll back to a previous revision. This is particularly useful when you experiment with prompt tweaks or refactor a data schema and need to revert to a known good state. The workspace-scoped auto-vault feature is also included in the free tier. Calling `vault_open_workspace(path)` creates a vault bound to a specific directory; calling it again with the same path returns the existing vault rather than creating a duplicate. This mirrors the workspace model most developers already use, so onboarding a new project is as simple as pointing LoreDocs at the project folder. Because the storage is file-based -- `loredocs.db` for the index plus a `vaults/` subdirectory for the raw documents -- your data is portable. Copy the directory to share with a teammate, archive it for compliance, or move it to a new machine. The data never leaves your infrastructure. For MCP-compatible clients, LoreDocs acts as a native MCP server for Claude Code, Cowork, the OpenAI Codex desktop app, Cursor IDE, and Hermes Agent. Configuration requires only a project-local `.mcp.json` (or `.cursor/mcp.json` for Cursor) and no additional setup. If your environment does not read `.mcp.json`, the `query_loredocs.py` Python fallback script lets you query any vault from automation scripts, CI pipelines, or Jupyter notebooks. Bulk import from Obsidian and Notion is included in free. Point `vault_import_dir` at any Obsidian vault root and LoreDocs walks the directory tree, extracts YAML front-matter tags, and creates matching vault entries. Notion import uses a separate `vault_import_notion` tool, which pulls specific pages and databases directly from the Notion API rather than walking a local export. This makes migration from an existing knowledge base a one-step operation rather than a manual copy-paste session. Together, these features give you a robust, searchable, and version-controlled knowledge base that lives entirely on your machine, with no account to create and no subscription to activate before you can start using it. ### Where the free tier ends The vault count caps at three. If you need more logical containers -- perhaps one per client, per model family, or per regulatory domain -- the free tier will block the fourth vault creation with a clear error. Semantic search is a Pro feature. Free tier search is keyword-based (FTS5), which handles most queries well, but when you are searching for something you remember the meaning of rather than the exact words, the hybrid semantic index in Pro -- a small English embedding model fused with BM25 scoring -- handles that case where keyword search cannot. Auto-discovered document relationships are also Pro-only. Pro automatically links related documents based on content similarity, making it easier to navigate a web of experiment logs, data dictionaries, and model cards as your vault grows. ### When Pro becomes a practical upgrade The inflection point is usually the vault cap. If you run three vaults comfortably and never wish you had a fourth, the free tier covers your workflow. When you start working across multiple clients or domains and the cap gets in the way, the $9/month Pro tier removes it entirely. Semantic search becomes valuable when your corpus is large enough that you cannot reliably recall the exact words you used in a document. A vault with several hundred entries from months of experiments will surface the relevant result on a vague query like "the run where we switched from batch to streaming" where keyword search returns nothing. Both tiers store your data in the same local SQLite format you own and can move, delete, or query directly. Upgrading does not change where the data lives or who controls it. ### Getting started LoreDocs is available from PyPI and the Anthropic marketplace. Install, run `loredocs init`, and your first vault is available immediately -- no signup, no email, no credit card. The full toolset is at [labyrinthanalyticsconsulting.com/tools](https://www.labyrinthanalyticsconsulting.com/tools). For a deeper look at how LoreDocs fits into a local AI knowledge stack, see the [LoreDocs deep dive](https://www.labyrinthanalyticsconsulting.com/blog/loredocs-deep-dive-technical-case) and the [local knowledge vault post](https://www.labyrinthanalyticsconsulting.com/blog/loredocs-local-knowledge-vault). Get posts like this delivered weekly: subscribe to [Dispatches from the Labyrinth](https://labyrinthanalytics.substack.com/subscribe?utm_source=blog&utm_medium=blog&utm_campaign=substack_subscribe). Interested in a lifetime deal on LoreDocs Pro? [Join the waitlist](https://www.labyrinthanalyticsconsulting.com/lifetime).