Competitor Web-Change Watch Agent with Firecrawl Monitor

What This Builds

This recipe builds an agent that watches a set of competitor pages and alerts you only when something meaningful changes — a price drop, a new SKU, a new job listing, a changelog entry. The hard parts of change monitoring (scheduling, snapshot storage, diffing, retrying webhooks, filtering out ads and timestamps) are handled by Firecrawl /monitor, so your agent only has to react to the diff.

You describe what to watch in plain English; Firecrawl checks the page on a schedule and fires a signed webhook with a structured diff (added / removed / changed) only when there’s a real change. Your agent wakes, summarizes the diff into a human sentence, and pushes it to Slack or Telegram. Because nothing fires when nothing changes, you spend tokens only on actual events.

Product Shape

This is an event automation, not a chat bot. The trigger is a real webhook from Firecrawl, the work is short and durable, and the output is a notification side effect. Keep webhook ingestion fast and run the summarization as a background task.

The Stack

• Firecrawl /monitor — scheduled checks on pages or whole sites, structured diffs, and signed webhooks that fire only on meaningful change. This is the entire monitoring stack in one endpoint.
• Trigger.dev — receives the Firecrawl webhook and runs the summarize-and-notify step as a durable task with retries, so a slow LLM call never drops an alert.
• An LLM such as Claude via the Anthropic Startup Program — turns the raw diff into a short, plain-English “what changed and why it matters” summary.
• Telegram Bot API (or Slack) — the delivery channel for the alert.

Step-by-Step Outline

1. Create a Firecrawl monitor for each competitor page (or a whole site), describing in plain English what to track and setting a cadence (hourly/daily/cron).
2. Point the monitor’s signed webhook at a Trigger.dev endpoint and verify the signature on receipt.
3. On a change event, enqueue a durable task instead of processing inline.
4. In the task, feed the structured diff to the LLM and ask for a one-line summary plus a “matters / ignore” classification.
5. Post meaningful changes to Telegram/Slack with the permalink to the diff; drop noise.
6. Add per-event subscriptions so the agent only wakes for the change types you care about.

Why This Shape Works

The classic build-it-yourself monitor is mostly plumbing — cron, storage, diffing, webhook retries, noise filtering — and Firecrawl /monitor collapses all of it into one endpoint that ingests up to 90% fewer tokens by sending only what changed. That leaves your agent to do the one thing it’s good at: judging whether a change matters and saying so clearly. Separating fast webhook ingress from durable summarization keeps alerts reliable.

Source

Firecrawl, Introducing /monitor: Notify AI agents when the web changes: https://www.firecrawl.dev/blog/firecrawl-monitoring-launch

Open-source companion — Firecrawl Observer (self-hostable monitor with visual diffs): https://www.firecrawl.dev/blog/introducing-firecrawl-observer