cbcoutinho/nextcloud-mcp-server

What it solves

This project provides a bridge between Large Language Models (LLMs) and Nextcloud instances. It allows AI assistants (like Claude, GPT, or Gemini) to interact with your personal or organizational data—such as notes, calendars, contacts, and files—using natural language commands instead of manual navigation.

How it works

It is a standalone Model Context Protocol (MCP) server. It connects to a Nextcloud instance via a secure API using app-specific passwords or OAuth (Login Flow v2). It exposes over 110 tools to AI clients, enabling them to perform CRUD operations across various Nextcloud apps (e.g., Notes, Calendar, Deck, Mail, and Talk).

Who it’s for

Users and administrators of Nextcloud who want to integrate their data into AI-powered workflows, as well as developers building AI assistants that need access to a self-hosted cloud storage and collaboration suite.

Highlights

• Broad App Support: Covers 10+ Nextcloud apps including Notes, Calendar, Contacts, Files, Deck, and Mail.
• Semantic Search: An experimental RAG pipeline using vector databases (Qdrant) and embedding services (Ollama) to find content by meaning rather than keywords.
• Document Processing: Built-in OCR and text extraction for PDFs, DOCX, and images.
• Flexible Deployment: Supports Docker, Kubernetes (via Helm), VM, or local installation.
• Secure Authentication: Implements Basic Auth with app passwords and a multi-user Login Flow v2.
• Tag-Based Exclusion: Ability to hide sensitive files or folders from the AI by using specific Nextcloud system tags.