Mcp Documentation Server
@andrea9293
MCP Documentation Server - Bridge the AI Knowledge Gap.
Overview
What is MCP Documentation Server?
A TypeScript-based Model Context Protocol (MCP) server that provides local-first document management and semantic search. Documents are stored in an embedded Orama vector database with hybrid search (full‑text + vector), intelligent chunking, and local AI embeddings — no external database or cloud service required.
How to use MCP Documentation Server?
Configure an MCP client (e.g., Claude Desktop) to run npx -y @andrea9293/mcp-documentation-server. The web UI starts automatically on port 3080. Use tools like add_document, search_all_documents, and process_uploads to manage and search documents. All environment variables are optional.
Key features of MCP Documentation Server
- Local-first document management and semantic search
- Hybrid full-text and vector search via Orama
- AI-powered search with optional Gemini API key
- Built-in web dashboard with drag-drop file uploads
- Parent‑child chunking for context‑preserving retrieval
- LRU embedding cache and streaming file reader
Use cases of MCP Documentation Server
- Indexing project documentation for instant semantic lookup
- Searching across code comments, markdown files, and PDFs
- Providing LLMs with richer context via neighboring chunk retrieval
- Managing a personal knowledge base without cloud dependencies
- Uploading and processing
.txt,.md, or.pdffiles in bulk
FAQ from MCP Documentation Server
Where is data stored?
All data resides in ~/.mcp-documentation-server/ (or a custom path via MCP_BASE_DIR). No external database or cloud service is used.
Is an external database or cloud service required?
No. The embedded Orama vector database runs locally; all embeddings are computed on‑device with Transformers.js.
Do I need a Gemini API key?
No. Without GEMINI_API_KEY, only local embedding‑based search tools are available. The AI‑powered search (search_documents_with_ai) requires the key.
Can I change the embedding model after adding documents?
Yes, but changing MCP_EMBEDDING_MODEL requires re‑adding all documents because embeddings from different models are incompatible. The Orama database is recreated automatically when the vector dimension changes.
How do I access the web UI?
The web UI starts automatically on port 3080 when the MCP server launches. Open http://localhost:3080. To run it standalone (without the MCP server), use npm run web or npm run web:build.
