RSS/Atom feed collection with 444 curated feeds. Python MCP server included.
FeedKit has 9 well-named tools with clear verb-noun patterns (fetch_, search_, collect_, subscribe_, unsubscribe_, list_, get_). All tools have descriptions (avg ~150 chars, within baseline 34-392 range). Input schemas are present with types and defaults. However, output schemas are NOT documented, tools return JSON strings without declaring field structure. Parameter descriptions are adequate but lack constraints (enums, ranges, format specs). Error handling is minimal; no recovery guidance or actionable error messages. Missing tool annotations (readOnlyHint, destructiveHint). No pagination support despite tools returning lists. Composition is sound, each tool has one responsibility and names are unambiguous.
Get statistics about the built-in feed catalog.
Collect new articles from all subscribed feeds. Fetches all subscriptions in parallel and stores new articles. Use search_feed_catalog to find feeds, then subscribe before collecting.
Fetch latest entries from a single RSS/Atom feed URL. Returns the most recent articles with title, URL, summary, and date. No subscription needed — just provide any RSS feed URL.
Get the most recently collected articles.
List all current feed subscriptions with their status.
Full-text search across all collected articles. Searches titles, summaries, and authors using SQLite FTS5. Articles must be collected first with collect_feeds.
Output schemas not documented. Tools return JSON strings (e.g., fetch_single_feed returns array of {title, url, summary, published, author}) but LLM cannot see field structure in tool definition. Forces LLM to parse unstructured output.
No pagination support. search_feed_catalog and search_articles accept 'count' but no offset/cursor. Large result sets will blow context window. Baseline pattern requires limit + pagination.
No tool annotations. Tools lack readOnlyHint (fetch_single_feed, search_feed_catalog, search_articles, get_latest_articles, list_subscriptions, catalog_stats) and destructiveHint (unsubscribe_feed). LLM cannot distinguish safe reads from destructive writes without reading descriptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
Search the built-in catalog of 444 curated RSS feeds. The catalog includes tech blogs (AWS, Cloudflare, Netflix...), academic journals, government sources, news outlets, and more. All feeds are verified working with 778K+ articles collected.
Subscribe to an RSS feed for ongoing collection.
Unsubscribe from a feed.
Parameter constraints missing. 'count' params lack min/max bounds (e.g., search_feed_catalog count defaults to 20 but no max specified). 'category' and 'language' params accept free-form strings instead of enums. LLM may pass invalid values.
No error handling guidance. Tools return JSON strings on success but no documented error responses or recovery hints. If collect_feeds fails on a feed, no indication whether to retry, skip, or ask user.