RSSMonster

WHAT IS IT?

RSSMonster is a self-hosted RSS reader built with Vue 3 and Express. It keeps the usual feed inbox and adds an organization layer on top: articles covering the same occurrence are grouped together, and a search query can become a permanent view. Reading happens in a three-panel layout, a stream of full articles or compact summaries, with keyboard shortcuts, touch gestures and a dark theme. It runs as a server through Docker Compose (SQLite or MySQL) or as an Electron desktop app for Windows, macOS and Linux.

WHY IS IT INTERESTING?

• Events and duplicates: Semantic Events gather articles about the same occurrence and list the sources that covered it. Duplicate detection is a separate step that tells a syndicated copy apart from another publisher's own reporting.
• Explainable ranking: the Recommended sort puts forward topics close to your interests, learned from your reading feedback (Interest Islands). Top Stories surfaces topics picked up by many sources. The signals behind a ranking can be inspected, and chronological order is still available.
• Smart folders and rules: a search like @today unread:true sort:desc can be saved as a Smart Folder that fills itself with new matching articles. Actions apply regular expressions to incoming articles to tag, favorite, mark read or hide them, with no AI involved.
• Local or remote AI: the MySQL profile runs ONNX models locally (Qwen for embeddings and generation, ModernBERT for scoring), with no OpenAI key required. The SQLite and desktop profiles ship without models but can connect to a remote inference service. The optional chat assistant uses an OpenAI provider.
• Works with what you already use: OPML import/export, Fever and Google Reader APIs for third-party clients, private generated RSS feeds from your selections, and a PWA with Web Push notifications. An MCP endpoint (/mcp), authenticated with a JWT, also exposes article search and retrieval tools to AI agents.

USE CASES

• Follow a busy news story without rereading the same wire report picked up by several outlets.
• Run a tech watch with one Smart Folder per project and an action that tags release announcements as they arrive.
• Keep a favorite Fever or Google Reader compatible client, with the self-hosted instance as its backend.
• Query archived articles from an AI agent through the MCP endpoint.
• Read feeds locally with the Electron app, without Docker or a server, refreshing feeds manually.