Agentic RSS Parser

Why it exists

rss-parser hasn't received a security update since 2022, yet it remains widely used across production agent pipelines. Agentic RSS Parser is a from-scratch rebuild that keeps the API you already know while closing the supply-chain gap — zero runtime dependencies, a custom XML engine, and native LLM adapters built directly on fetch().

Migrate in one line

All existing parseURL, parseString, parseFile, customFields, headers, timeout, and callback-style usage is preserved exactly. The agentic pipeline is an optional layer on top.

// Before
import Parser from 'rss-parser';

// After — zero other changes needed
import Parser from 'agentic-rss-parser';

Core features

• RSS 2.0, Atom & JSON Feed — namespaces, CDATA, HTML entities, dc:creator, media:content, content:encoded, plus native JSON Feed v1/v1.1 detection and parsing
• Drop-in compatibility — parseURL, parseString, parseFile, customFields, callback and promise styles
• Podcast & media enclosures — <enclosure> audio/video metadata, media:thumbnail / media:content, and iTunes podcast tags normalised on every item
• OPML import — parseOpml() extracts feed URLs and categories from OPML subscription lists, including nested folders
• Continuous feed watcher — createFeedWatcher() is an event emitter for background polling with automatic conditional-GET support and result / poll / feedError events
• Persistent HTTP caching — ETag / Last-Modified are stored per feed, so repeat polls send conditional GETs and skip re-analysing unchanged feeds on a 304
• Vector database export — storage.exportForEmbedding() formats analysed items into ready-to-embed documents for Pinecone, Chroma, pgvector, or any other vector store
• Search & analytics APIs — storage.searchAnalyses() for keyword search across stored intelligence, storage.getStatistics() for database-wide metrics
• Heuristic analysis — signal-based relevance scoring, no API key needed, fully configurable via signals, extraSignals, and threshold
• LLM analysis — OpenAI, Anthropic, and local (Ollama) providers via native fetch(), with custom systemPrompt / promptTemplate support for domain-specific evaluation
• Article enrichment — fetches and strips full article body from feed item URLs
• MCP server — stdio JSON-RPC 2.0 server exposing five tools (feed fetch, article fetch, history search, statistics, database pruning) plus a live resource endpoint, compatible with Claude Desktop, Cursor, VS Code, and any MCP host
• SQLite deduplication — items are SHA-256 deduplicated across runs, with pruneOlderThan() for long-running deployments
• Retry & conditional GET — automatic exponential-backoff retries on 429/5xx, plus ETag / If-Modified-Since support to skip unchanged feeds
• SDK integrations — ready-to-use examples for Anthropic SDK, OpenAI Agents SDK, Vercel AI SDK, LangChain, and Google ADK
• userAgent override — avoid 403s on feeds that block bot user-agents (Reddit, HN, Lobste.rs)

MCP tools & resources

Exposed over stdio for Claude Desktop, Cursor, and Cline:

• fetch_rss_feed — fetch and agentically analyse an RSS, Atom, or JSON feed; supports heuristic and LLM providers
• fetch_full_article — fetch and HTML-strip an article body for context
• search_feed_history — search stored intelligence and analyses by keyword
• get_feed_statistics — retrieve metrics on processed feeds and relevance ratios
• prune_database — prune items older than a given TTL in days
• rss://analyses/latest — MCP resource for reading recent relevant intelligence straight from the SQLite database

New in v1.7.0

Security hardening and feature expansion: credential isolation so environment API keys are never auto-forwarded to a custom MCP baseURL, prototype-pollution and XML entity/nesting DoS hardening in the parser, complete IANA CIDR range coverage for SSRF blocking (multicast, reserved, documentation, benchmarking, and 6to4 ranges), native JSON Feed support, podcast/media enclosure normalisation, persistent ETag-based HTTP caching, and three new MCP tools plus a live resource endpoint.

Quick start

npm install agentic-rss-parser

import Parser from 'agentic-rss-parser';

const parser = new Parser({ timeout: 10000 });
const feed = await parser.parseURL('https://news.ycombinator.com/rss');

for (const item of feed.items) {
  console.log(`- ${item.title} — ${item.link}`);
}

Agentic pipeline

Run the full fetch → parse → deduplicate → analyse pipeline with zero API key required, using the built-in heuristic engine:

import { runAgenticParser } from 'agentic-rss-parser';

const { results, feedErrors } = await runAgenticParser({
  feedUrls: ['https://news.ycombinator.com/rss'],
  dbPath: './data/rss-agent.db'
});

for (const { item, analysis } of results) {
  if (analysis.decision === 'relevant') {
    console.log(`[${analysis.confidence}%] ${item.title}`);
  }
}

Import an OPML list and watch it continuously

Import a subscription list and poll it in the background without wiring a timer yourself.

import { parseOpml, createFeedWatcher } from 'agentic-rss-parser';
import { readFileSync } from 'node:fs';

const { feeds } = parseOpml(readFileSync('./subscriptions.opml', 'utf8'));

const watcher = createFeedWatcher({
  feedUrls: feeds.map((f) => f.xmlUrl),
  intervalMs: 5 * 60_000,
  dbPath: './data/rss-agent.db'
});

watcher.on('result', ({ item, analysis }) => {
  if (analysis.decision === 'relevant') console.log(item.title);
});

watcher.start();

Requirements

• Node.js ≥ 22.5.0 (required for node:sqlite)
• ESM-only package
• Linux, macOS, Windows