B2B SaaS Changelog & Feature Intelligence Monitor

Track any SaaS competitor's changelog automatically. This Actor extracts, normalizes, and diffs changelog / release-note data from any SaaS URL — RSS feeds, Next.js apps, HeadwayApp, Canny, Beamer, GitHub Releases, or plain custom HTML — and outputs clean, LLM-ready Markdown plus structured JSON. Only genuinely new entries since your last run are flagged, so you can pipe results straight into Slack, Notion, or a n8n / Make.com workflow without re-processing duplicates.

Why use this Actor?

• Works on almost any changelog, out of the box. A strict extraction waterfall tries the highest-fidelity method first and falls back gracefully:
  1. RSS / Atom / JSON feed discovery
  2. Next.js __NEXT_DATA__ hydration payload
  3. Platform adapters — HeadwayApp, Canny, Beamer, GitHub Releases
  4. Generic Readability-based fallback for any other custom HTML page
• Stateful diffing, no duplicate noise. Every entry gets a stable content hash stored in a Key-Value Store across runs, so scheduled runs only report what's actually new.
• LLM-ready output. Every entry ships as clean Markdown (HTML boilerplate stripped) with an estimated token count, ready to drop into a prompt, RAG pipeline, or AI agent.
• Zero anti-bot headaches. Changelogs are public pages — this Actor uses a fast, lightweight CheerioCrawler (HTTP-only), no headless browser required.
• Auto-categorized. Each entry is tagged as New Feature, Improvement, Bug Fix, Breaking Change, Security, or General Update based on content heuristics.

Use cases

• Competitive intelligence — monitor competitors' changelogs and get notified the moment they ship a feature that matters to you.
• Product management — keep a single feed of everything shipped across your own product's multiple changelog sources (marketing site + docs + GitHub).
• AI workflows — feed new releases into an LLM to auto-draft competitor battlecards, sales talk tracks, or internal Slack digests.
• VC / market research — track feature velocity across a portfolio of SaaS companies over time.

Input

Field	Type	Required	Description
startUrls	array of strings	✅	Absolute http(s) URLs to changelog, release-notes, or GitHub Releases pages (e.g. https://linear.app/changelog, https://github.com/vercel/next.js/releases).
onlyNewSinceLastRun	boolean	–	When true (default), only entries not seen in a previous run are included in the output. Set false to always emit the full current snapshot.
maxItemsPerSource	integer	–	Optional cap on how many entries to keep per source URL (most recent first). Leave empty for no limit.
requestTimeoutSecs	integer	–	Per-URL timeout in seconds for fetching the page and any discovered feed. Default 60.
proxyConfiguration	object	–	Optional Apify Proxy or custom proxy URLs.

Example input

{
  "startUrls": [
    "https://linear.app/changelog",
    "https://github.com/vercel/next.js/releases",
    "https://supabase.com/changelog"
  ],
  "onlyNewSinceLastRun": true,
  "requestTimeoutSecs": 60
}

Output

One dataset record per changelog entry (not per source URL) — the actual scraped content, flat and ready to consume:

{
  "sourceUrl": "https://linear.app/changelog",
  "hostname": "linear.app",
  "status": "OK",
  "extractionMethod": "rss",
  "isNew": true,
  "scrapedAt": "2026-08-27T04:10:32.123Z",
  "id": "b2c1e4f9a6d7...",
  "title": "Improved keyboard navigation in the Inbox",
  "contentMarkdown": "We've reworked keyboard shortcuts across the Inbox...\n\n- Jump between threads with J/K\n- Archive with E\n- Snooze with H",
  "features": [
    "Jump between threads with J/K",
    "Archive with E",
    "Snooze with H"
  ],
  "url": "https://linear.app/changelog/2026-08-20-inbox-keyboard-nav",
  "publishedAt": "2026-08-20T00:00:00.000Z",
  "category": "Improvement",
  "tags": [],
  "estimatedTokens": 84
}

• contentMarkdown — the full release body/content (from the RSS <content:encoded>/<description>, Atom <content>/<summary>, or the matched adapter's HTML), converted to clean Markdown.
• features — the top bullet-point lines pulled out of contentMarkdown, for a quick scannable summary without parsing the full body yourself.
• isNew — whether this entry wasn't seen in any previous run (see onlyNewSinceLastRun below).

A source URL that fails to process still produces exactly one record, so a single bad URL never crashes the whole run and the failure stays visible in the dataset:

{
  "sourceUrl": "https://example.com/broken",
  "hostname": "example.com",
  "status": "FAILED",
  "error": "Request timed out",
  "scrapedAt": "2026-08-27T04:10:32.123Z"
}

Pricing

This Actor uses Apify's Pay-Per-Event model — no flat platform subscription:

Event	Price	When it's charged
Base run cost	$0.005	Once per Actor run
Per URL processed	$0.005	Per source URL crawled
Per new changelog detected	$0.02	Per genuinely new entry found (skipped for duplicates)

Integrations

Connect the dataset output to:

• Slack / Discord — via Apify's built-in webhook integrations, alert your team the moment a competitor ships something new.
• n8n / Make.com — trigger a scenario on new dataset items to auto-summarize with an LLM and post to Notion, Airtable, or a CRM.
• Zapier — use the Apify Zapier integration to route new entries into any downstream app.

Example recipe: Automated Slack Digest (Make.com / n8n)

Turn this into a daily "what did our competitors ship" digest with zero manual work:

1. Schedule the Actor — in the Apify Console, open this Actor → Schedules tab → create a new schedule (e.g. 0 8 * * * for 08:00 AM UTC daily) with your list of competitor startUrls.
2. Keep onlyNewSinceLastRun: true in the schedule's input — this ensures the run only reports items that weren't already seen in a previous run, so your digest never repeats itself.
3. Add a webhook on run completion — Actor → Integrations → Webhooks → create one for the ACTOR.RUN.SUCCEEDED event, pointing at:
   • a Make.com scenario's "Custom Webhook" trigger, or
   • an n8n workflow's "Webhook" node.
4. Fetch the dataset in your scenario/workflow — the webhook payload includes resource.defaultDatasetId; use Apify's "Get Dataset Items" module/node (or a plain GET https://api.apify.com/v2/datasets/{datasetId}/items) to pull the new records.
5. Filter for isNew: true so the flow (and your Slack channel) stays silent on days with no changes.
6. Format and post to Slack/Discord — each dataset record is already one changelog entry, so post one message per record (or group them into one digest) to your team's #competitor-intel channel, including title, category, url, and the features bullet list or full contentMarkdown.

This gives you a fully automated, noise-free "competitor changelog digest" landing in Slack every morning with $0 infrastructure to maintain.

Model Context Protocol (MCP) Setup

You can also call this Actor as a live tool directly from an AI agent like Claude Desktop or Cursor, via Apify's MCP server. Once connected, you can just ask your agent things like "Check linear.app/changelog for anything new" and it will run this Actor and read the results back into the conversation.

Option A — Hosted remote server (recommended, no install)

No npx, no local Node process, and it auto-updates. Add a custom MCP connector pointing at:

https://mcp.apify.com

• Claude Desktop: Settings → Connectors → Add custom connector → paste the URL above → Connect (authorizes via OAuth, no token to copy/paste).
• Cursor: Settings → MCP → Add new MCP server → paste the URL above.

Option B — Local stdio server

Add this to claude_desktop_config.json (Claude Desktop) or .cursor/mcp.json (Cursor):

{
  "mcpServers": {
    "apify": {
      "command": "npx",
      "args": [
        "-y",
        "@apify/actors-mcp-server",
        "--actors",
        "dataflow-tools/saas-changelog-feature-intelligence-monitor"
      ],
      "env": {
        "APIFY_TOKEN": "YOUR_APIFY_API_TOKEN"
      }
    }
  }
}

• Get YOUR_APIFY_API_TOKEN from Apify Console → Settings → Integrations.
• The --actors value whitelists which Actor(s) show up as tools — swap in a comma-separated list to expose more than one.
• Fully quit and reopen Claude Desktop / Cursor after saving the config for it to pick up the new server.

Note: this only works with your own APIFY_TOKEN while the Actor is private. To let other people reference dataflow-tools/saas-changelog-feature-intelligence-monitor with their own tokens, publish the Actor to the Apify Store first (Console → Actor → Publication tab).

How it works (extraction waterfall)

1. RSS/Atom/JSON feed discovery — looks for <link rel="alternate"> tags and common feed paths (/feed, /rss.xml, /changelog.rss, etc.).
2. Next.js hydration data — parses <script id="__NEXT_DATA__"> for server-rendered pageProps containing changelog entries.
3. Platform adapters — dedicated parsers for HeadwayApp, Canny, Beamer widgets, and GitHub Releases (converted to their Atom feed).
4. Generic fallback — uses Mozilla's Readability algorithm to extract the main article content from any other page, then splits it into entries by heading.

Whichever method succeeds first is used — later stages are only attempted if earlier ones return no results.

FAQ

Does this work on Canny/Headway/Beamer widgets even if they're client-side rendered? For platforms that render primarily via JavaScript (e.g. some Beamer embeds), the Actor extracts whatever is present in the initial server-rendered HTML or inlined script payload. If a source renders 100% client-side with no fallback data, the generic Readability extractor is used as a last resort.

Can I track only new items and never see duplicates again? Yes — leave onlyNewSinceLastRun at its default (true). The Actor persists a stable hash per entry across runs in a named Key-Value Store, so re-running on a schedule only surfaces what's new.

Can I run this on a schedule? Yes — set up an Apify Schedule against this Actor (e.g. daily or hourly) with the same input, and combine it with a webhook to get notified only when newItemCount > 0.

What if a source URL is temporarily down? That source's record is marked "status": "FAILED" with an error message; all other URLs in the same run are processed normally.