About this project

Mediabox MCP is presented as a self-hosted media server stack whose management layer is exposed through the Model Context Protocol (MCP). The repository describes three ways to run it: a Tauri desktop application for Windows/macOS and local-first installs; an interactive CLI wizard, npx create-mediabox, intended for Linux servers, VPS, and headless deployments; and a headless MCP server that can be connected to Claude Desktop, ChatGPT, Gemini, an OpenAI-compatible client, or an optional Telegram bot over OAuth-protected Streamable HTTP. All three surfaces share the same Docker stack, the @mediabox/core orchestration pipeline, and the same MCP tools. Deployment modes include Local for home networks, VPS with Caddy and automatic HTTPS, and Cloudflare Tunnel for public access without opening ports. Requirements stated are Docker, Docker Compose, and Node.js 20 or newer. Desktop builds additionally need Rust for Tauri and Bun to compile the Node sidecar. The CLI supports flags such as --generate-only to write configuration without starting Docker, while --local-build is aimed at contributors running from a cloned repository. The stack centers on Jellyfin, Sonarr, Radarr, qBittorrent, and PyLoad, with Prowlarr and FlareSolverr also shown in the architecture. A reverse proxy such as Caddy or nginx fronts the services. The MCP server exposes /mcp, /api/dashboard, /api/chat, and /api/setup, and the README lists 30 MCP tools grouped into Jellyfin, Library, Sonarr, Radarr, Downloads, and Maintenance categories. Examples include server status and activity logs, media search and details, library and file management, subtitle fixing, batch renaming, series and movie search/status/removal/releases/grab/import/rescan, direct URL and PyLoad downloads, download queue management, and media optimization or cleanup jobs. The setup flow is described as replacing roughly fifteen manual steps. It asks for deployment mode, media paths, credentials, timezone, and optional integrations; generates .env, docker-compose.yml, a Caddyfile for VPS, and qBittorrent configuration; starts Docker containers; then auto-configures the stack through service APIs. This includes extracting Sonarr/Radarr/Prowlarr API keys, running the Jellyfin setup wizard, creating an admin user and API key, configuring qBittorrent as a download client, adding root folders, syncing Prowlarr indexers, setting up FlareSolverr and Jellyfin media libraries, and setting web UI credentials. The repository notes that adding torrent indexers in Prowlarr remains a manual step. The project is MIT licensed and the README is available in English and Spanish.