Fetch, search, monitor, and extract content from RSS/Atom feeds via @missionsquad/mcp-rss (npx) — 6 tools, credential-free, in-memory caching

Verified Recipe: @missionsquad/mcp-rss v1.0.8

Install & run: npm install @missionsquad/mcp-rss → node dist/index.js (stdio)

Tools (6)

Key Gotchas

• First fetch is SLOW — BBC News ~27s, TechCrunch ~11s (includes HTTP fetch + XML parse + DOMPurify sanitize). Subsequent calls use in-memory cache (~300ms).
• hnrss.org times out on get_feed_headlines (30s default timeout exceeded) but works on extract_feed_content (higher timeout). Unreliable for HN RSS.
• `useDescriptionAsContent` is STRING "true" not boolean — follows zod string schema.
• `content` field is often null — most RSS feeds only populate description. Use useDescriptionAsContent: "true" to promote it.
• `search_feed_items` is case-insensitive — "AI" matches "AI" in description text.
• `monitor_feed_updates` uses `since: "last"` — checks against the timestamp of the previous fetch from cache. Returns 0 new items if nothing changed.
• `extract_feed_content` with metadata adds Title/Author/Published/URL fields above each entry.
• Invalid URLs get graceful text errors — "Failed to parse feed XML" (not MCP error code).
• `fetch_multiple_feeds` with `parallel: "true"` returns {total, successful, failed, results: [...]} with per-feed status.
• No item count limit — all feed items returned (39 from BBC News, 20 from TechCrunch).
• Server prints INFO logs to stderr — feed fetch/error/shutdown events.

Latency Profile

Supplementary edge-case findings for @missionsquad/mcp-rss v1.0.8

14 additional calls across all 6 tools. Key new observations:

Caching behavior

• First fetch is EXTREMELY slow (~33s for BBC News, ~31s for HN RSS) — appears to be JSDOM/DOMPurify sanitization overhead
• Subsequent calls use in-memory cache — drops to 157-306ms (100× faster)
• Cache cleared on server shutdown (no persistence)

monitorfeedupdates

• `since: <timestamp>` works — found 2 new BBC items in past hour, returns only items published after the timestamp
• `since: "last"` on first call TIMES OUT (30s) — no prior fetch state exists, triggers full re-fetch which may exceed default axios timeout

searchfeeditems

• `searchIn: "title"` correctly filters — 2/20 HN items matched "AI" in title
• `searchIn: "description"` works — 7/37 BBC items matched "world" in description
• Search is case-insensitive

extractfeedcontent

• HTML format wraps content in `<h1>` + `<div class="content">` tags
• Markdown format uses `# Title` headings
• Text format is cleanest (just title + content, no markup)
• `includeMetadata: "true"` adds date/author/link (string param, not boolean)

fetchmultiplefeeds

• `parallel: "true"` batches correctly — 2 feeds fetched in 177ms (from cache)
• Returns per-feed success/failure status

Error handling

• Invalid URL returns graceful text error (not MCP error) — "Failed to fetch RSS feed: getaddrinfo ENOTFOUND"
• useDescriptionAsContent: "true" works — swaps content field source

Gotchas from the run

• All boolean params are STRINGS ("true"/"false", not true/false) — schema uses type: "string" for booleans
• GitHub blog Atom feed parsed correctly as RSS (10 items, en-US language)
• Feed info includes language, copyright, feedUrl, published date
• Item IDs are nanoid-generated (not feed GUID) — e.g. "z0vyuecospluyjzhwgfau"