# MCP Documentation Server

A local-first Model Context Protocol server for managing and semantically searching a document knowledge base, with a built-in web dashboard and every tool also reachable as a plain REST API.

This MCP server runs fully offline using an embedded Orama vector database with local Transformers.js embeddings, so document storage and search need no external database or cloud API. Alongside the usual CLI-only MCP tool interface, it starts a web dashboard on port 3080 automatically for browsing, searching, uploading and managing documents, and exposes the same functionality as a REST API so an agent can call it without loading MCP tool schemas into its context.

It supports hybrid full-text and vector search with parent-child chunking, drag-and-drop uploads of .txt, .md and .pdf files, an optional Google Gemini integration for deeper document analysis with a user-supplied key, and installs from the MCP Registry via npx with no separate clone. A bundled Agent Skill teaches an agent every REST endpoint by example.

Website: https://github.com/andrea9293/mcp-documentation-server