FeedKit

FeedKit

[PyPI] [Python] [License] [Tests]

한국어 문서 · llms.txt

RSS/Atom feed collection with 444 curated, verified feeds. CLI + Python API + MCP server.

Quick Start

pip install feedkit
feedkit search cloudflare           # search the built-in catalogfeedkit subscribe-catalog -c technology   # subscribe to all 68 tech feedsfeedkit collect                     # fetch all subscriptions (async parallel)feedkit find "kubernetes"           # full-text search collected articles

Install

pip install feedkit             # core (CLI + Python API)pip install "feedkit[mcp]"      # + MCP serverpip install "feedkit[all]"      # + MCP + OPML import/export

Requirements: Python 3.11+

How It Works

flowchart LR    A["📋 Catalog\n444 curated feeds"] --> B["Subscribe\nselect feeds"]    B --> C["🔄 Collect\nfetch new entries"]    C --> D["📦 Store\nSQLite database"]    D --> E["🔍 Query\nfilter & retrieve"]

CLI Reference

Catalog Commands

feedkit search [QUERY]

Search the built-in catalog of 444 curated feeds.

feedkit search aws                    # search by title or domainfeedkit search --category science     # filter by categoryfeedkit search --language ko          # filter by languagefeedkit search -c finance -l en -n 50 # combine filtersfeedkit search -j                     # JSON output (for piping)

feedkit categories

List all available catalog categories.

$ feedkit categories  academia  finance  pets  science  society  technology

feedkit stats

Show catalog and local subscription statistics.

$ feedkit statsCatalog: 444 feeds  academia: 13  finance: 89  pets: 27  science: 128  society: 119  technology: 68
Local: 68 subscriptions, 1,247 articles

Subscription Commands

feedkit subscribe <URL>

Subscribe to a single feed.

feedkit subscribe https://blog.cloudflare.com/rss/feedkit subscribe https://example.com/rss -c tech -t "My Feed"

feedkit subscribe-catalog -c <CATEGORY>

Subscribe to all feeds in a catalog category at once.

feedkit subscribe-catalog -c technology   # subscribe to all 68 tech feedsfeedkit subscribe-catalog -c science      # subscribe to all 128 science feeds

feedkit unsubscribe <URL>

Remove a feed subscription and its collected articles.

feedkit list

List all current subscriptions with fetch counts and error counts.

Collection Commands

feedkit collect

Fetch new articles from all subscribed feeds (async parallel).

feedkit collect                    # collect allfeedkit collect -c technology      # collect only tech feedsfeedkit collect -n 50              # max 50 concurrent requests

Output:

Collecting from 68 feeds... 67/68 feeds OK, 412 new articles, 8234ms

feedkit latest

Show most recently collected articles.

feedkit latest                # latest 20 articlesfeedkit latest -n 50          # latest 50feedkit latest -c finance     # latest from finance only

feedkit find <QUERY>

Full-text search (SQLite FTS5) across all collected articles.

feedkit find "kubernetes deployment"feedkit find "large language model" -n 50

OPML Commands

feedkit import-opml <PATH>

Import feeds from an OPML file (Feedly, Inoreader, NetNewsWire, etc.).

feedkit import-opml subscriptions.opml

feedkit export-opml <PATH>

Export current subscriptions to OPML.

feedkit export-opml backup.opml

Python API

from feedkit import search_catalog, fetch_feed, get_catalog_stats, FeedStorefrom feedkit.core import collect
# Search the built-in catalogfeeds = search_catalog("cloudflare")feeds = search_catalog(category="technology", language="en", limit=50)
# Fetch a single feed (async)entries = await fetch_feed("https://blog.cloudflare.com/rss/")for entry in entries:    print(entry.title, entry.url, entry.published)
# Subscribe and collectstore = FeedStore()                              # SQLite at ~/.feedkit/feedkit.dbstore.subscribe("https://blog.cloudflare.com/rss/", category="tech")result = await collect(store, concurrency=20)    # async parallel fetchprint(f"{result.new_articles} new, {result.feeds_ok}/{result.feeds_total} OK")
# Search collected articles (FTS5)articles = store.search("kubernetes", count=10)
# Latest articlesarticles = store.get_latest(count=20, category="tech")
# Feed healthhealth = store.get_health()                      # fetch/error counts per feed
# Catalog statsstats = get_catalog_stats()                      # {total_feeds, categories, languages}
store.close()

Key Classes

MCP Server

pip install "feedkit[mcp]"feedkit-mcp                    # starts stdio MCP server

Configuration

Claude Code (~/.claude/settings.json):

{  "mcpServers": {    "feedkit": {      "command": "feedkit-mcp"    }  }}

Tools Reference

MCP Workflow Example

User: "What are the latest AI papers on arXiv?"
1. search_feed_catalog(query="arxiv", category="science")   → finds arXiv cs.AI, cs.LG, cs.CL feeds
2. subscribe_feed(url="https://rss.arxiv.org/rss/cs.AI")
3. collect_feeds(category="science")   → fetches new entries
4. get_latest_articles(category="science", count=10)   → returns latest papers

Built-in Catalog

444 verified feeds across 6 categories. All audited — hard paywalls (Bloomberg, FT, WSJ) and broken URLs removed.

Languages: English (381), Korean (47), Japanese (14), Chinese (2)

Full feed list: CATALOG.md

Why FeedKit?

Disclaimer

This package distributes a catalog of publicly available RSS feed URLs, not the feed content itself. RSS is a syndication standard — publishing an RSS feed is an explicit invitation for readers to subscribe. FeedKit fetches feeds on the user's behalf and does not store or redistribute copyrighted content.

Feeds behind hard paywalls (Bloomberg, FT, WSJ, Barron's) and feeds with aggressive terms of service have been removed from the catalog.

Used in

• newswatch — RSS news monitoring pipeline (feedkit → markgrab → embgrep → diffgrab)

License

MIT

_{Part of the QuartzUnit ecosystem — composable Python libraries for data collection, extraction, search, and AI agent safety.}