Rag Vault
Give your coding agent a local knowledge base with hybrid search over PDFs and DOCX without standing up a separate vector stack.
Overview
RAG Vault is an MCP server for the Build phase that provides local hybrid RAG search over PDF and DOCX files for AI coding agents.
What is this MCP server?
- Local RAG with hybrid search (semantic plus keyword via RAG_HYBRID_WEIGHT, default 0.6)
- Ingest PDF and DOCX with LanceDB storage and Transformers.js embeddings
- Zero-config defaults: Xenova/all-MiniLM-L6-v2, ./lancedb/, configurable BASE_DIR
- npm package `@robthepcguy/rag-vault` v1.5.3 with stdio transport
- Tune MODEL_NAME, DB_PATH, CACHE_DIR, and RAG_GROUPING without cloud dependency
- Registry version 1.5.3
- Default embedding model Xenova/all-MiniLM-L6-v2
- Default hybrid keyword weight 0.6 per RAG_HYBRID_WEIGHT
Community signal: 1 GitHub stars.
What problem does it solve?
Solo builders want agent memory over private docs but do not want to operate a separate cloud vector database during development.
Who is it for?
Indie developers who keep specs and manuals locally and need MCP retrieval in Claude Code or Cursor without SaaS RAG bills.
Skip if: Teams that require multi-tenant cloud RAG, heavy OCR pipelines, or enterprise compliance hosting out of the box.
What do I get? / Deliverables
After you register RAG Vault, your agent queries a on-machine LanceDB index with hybrid search instead of hallucinating over unread PDFs.
- Indexed local corpus searchable via MCP tools
- Hybrid retrieval tuned with RAG_HYBRID_WEIGHT and optional RAG_GROUPING
- Agent answers grounded in your PDF and DOCX sources
Recommended MCP Servers
Journey fit
RAG vault is build-time infrastructure that makes agents smarter about your repo docs, specs, and uploads while you implement features. Agent-tooling is the canonical shelf because the server exists to extend MCP agents with retrieval, not to ship end-user frontend UI.
How it compares
Self-hosted document RAG MCP, not a hosted Pinecone connector or single-shot summarization skill.
Common Questions / FAQ
Who is RAG Vault for?
RAG Vault is for solo builders and small teams who want private, local document search inside MCP agents while building products.
When should I use RAG Vault?
Use it during Build when you have PDFs, DOCX, or internal docs the agent must cite while implementing features or debugging integrations.
How do I add RAG Vault to my agent?
Install `@robthepcguy/rag-vault` from npm, add the stdio MCP server in your host, set BASE_DIR and optional DB_PATH or MODEL_NAME, then ingest documents through the server tools.