What Is This?

PageCrawl ships a Model Context Protocol (MCP) server that plugs directly into Claude. Once it is connected, you can describe what you want to monitor in plain English and Claude makes the API call for you, picking the right tracking mode, selector, frequency, and notification channel from your sentence. No tab-switching, no dashboard form.

This guide assumes you already have the MCP server connected. If you don't, the MCP integration guide covers setup for Claude.ai web, Claude Code, Cursor, and other MCP-compatible clients in five minutes.

Step 1: Make Sure Your Plan Allows Writes

Read tools (list-monitors, get-changes-since, get-check-diff) work on every plan, including Free. Creating new monitors through Claude requires Standard or above, because add-page-monitor is a write tool.

If you are on Free, Claude will return a clear error pointing you to the upgrade page. Read-only exploration of existing monitors still works.

Step 2: Describe What You Want to Monitor

Open Claude (web or CLI) and tell it what to watch. A few real prompts that work first-try:

Price tracking:

Watch https://www.apple.com/shop/buy-mac/macbook-pro and tell me when the price changes.

Claude calls add-page-monitor with tracking_mode: "price". The monitor tracks both the price and stock availability, runs daily, and notifies through your default channels.

Legal page monitoring:

Send a weekly Slack message if Anthropic's consumer terms change at https://www.anthropic.com/legal/consumer-terms.

Claude picks tracking_mode: "reader" (which strips nav and footer noise), sets frequency: 10080 for weekly, and routes notifications to Slack.

API monitoring:

Track the timestamp of the latest CVE published in https://services.nvd.nist.gov/rest/json/cves/2.0?resultsPerPage=1 and ping me hourly.

Claude picks tracking_mode: "json_path" and writes the JSONPath expression $.vulnerabilities[0].cve.published from your description.

Bulk setup:

Set up daily price-page monitors for these six competitor URLs and tag them all competitor-pricing: [list of URLs]

Claude calls add-page-monitor six times with consistent settings and tags. Six form submissions in the dashboard becomes one chat message.

Step 3: Verify the Monitor

After Claude creates a monitor, ask:

Show me the monitor you just created and confirm the first check ran.

Claude calls get-monitor-details and get-latest-values and reads back the configuration and the first captured value. If something looks off (wrong selector, wrong frequency, the value field is empty), correct it in the same conversation: "actually use ai_extract with the prompt 'find the headline price'" or "drop the frequency to 15 minutes".

Examples of Follow-Up Prompts

Once monitors are in place, the read-side prompts are where the integration earns its keep:

• Daily triage: "What changed across all monitors in the last 24 hours, sorted by priority?"
• Investigate a specific change: "Show me the exact diff for the monitor on stripe.com pricing."
• Force a recheck: "Trigger an immediate recheck on every monitor tagged competitor-pricing."
• Mark as reviewed: "Mark everything as seen, I am caught up."

Each of these maps to one or two MCP tool calls (get-changes-since, get-check-diff, trigger-check, mark-changes-seen) that Claude picks automatically.

Claude.ai Web vs Claude Code

Both surfaces talk to the same MCP server with the same tools. The right one depends on how you work:

• Claude.ai web: best for one-off setup, conversational triage, and sharing results with a teammate via a chat link. OAuth handshake the first time, then nothing to maintain.
• Claude Code: best when monitor setup is part of a development workflow. Token-based authentication makes it scriptable. Pipe a CSV of URLs into a single Claude Code session and have it create every monitor with consistent tags, or wire the MCP server into a CI job that creates a monitor for every new deployment URL.

What This Doesn't Do

Some Advanced Setup features still belong in the web dashboard, not in chat:

• Authentication flows. If a page needs a login, configure the credentials in Settings → Authentications and reference the saved authentication from the dashboard. Sharing credentials over chat is a bad pattern, and the MCP server intentionally does not expose authentication management.
• Complex action sequences with custom JavaScript. Claude can configure clicks, waits, scrolls, and short JavaScript snippets through MCP, but for monitors that need multi-step interaction with framework-aware event dispatching, the dashboard's Advanced Setup editor (with autocomplete and syntax highlighting) is the better surface.

Tips

• Lead with the page URL. "Watch https://… and …" almost always picks the right tracking mode. Putting the URL first gives Claude the strongest signal.
• Name the channel explicitly when you want one. "Ping me on Slack" or "email only" overrides the workspace default.
• Use tags for groups you will query later. "Tag this competitor-pricing" makes future "show me changes in this group" prompts trivial.
• Correct in the same turn. Claude does not need a perfect prompt. "Use ai_extract instead, with the prompt 'find the EU launch date'" works as a follow-up.

Want the Deeper Tour?

The Setting up PageCrawl monitors by talking to Claude (web and CLI) via MCP blog post walks through five worked examples in detail, including the JavaScript tracked element and the actions array for monitors that need a click before extraction. Each example also shows the equivalent REST API call if you would rather script it from a shell.