Build an n8n competitor monitoring workflow from measured signals

Most n8n competitor monitoring templates scrape a page and compare it with an earlier copy. n8n's own template runs a loop of daily checks: it scrapes a page with Firecrawl, waits one day, scrapes again, and has an AI agent decide whether to send an email through Gmail. A template store describes the common alternative, which hashes the relevant part of a page, stores a baseline in Google Sheets and sends a Slack or email alert when the hash changes.

Both approaches share limits that the template store itself lists. Hashing a whole page produces false alerts from cookie banners, scripts and rotating content. Sites can rate limit requests. A hash change shows that something changed, not what, so an AI summary has to fill in the context. And the first run has no baseline, so it must be handled as a seed.

These workflows watch pages you point them at. They do not tell you what a competitor's changelog, blog or review profile said this week, in dated and graded form. That is the gap measured signals fill.

Linkeddit's competitor signals arrive already dated, graded and explained. On the test project, one call returned 19 signals for the week of 2026-10-26. Each has a priority, a category and a why_it_matters line, so the workflow's Switch can route two high signals to a channel and leave eight low ones in a log.

That removes the two jobs n8n scraping templates spend most of their nodes on: detecting a change and describing it. The workflow keeps the job it is good at, which is routing, de-duplicating and logging. Each signal also carries a source_url, so a person reading the alert can open the original post.

A competitive intelligence founder described the usual failure on Reddit: "Teams collect screenshots and links, put them into Notion, send a Slack update and move on. Nobody clearly owns the response." A routed alert with a priority and a named channel gives the response an owner.

n8n runs the workflow once it is published. Claude does not run on a timer, so the Linkeddit data reaches n8n when you start Claude and it writes the week's signals into the workflow's data table. That split is deliberate. n8n's own comparison of Claude Code and n8n says recurring, unattended processes with retries, logs and version history belong in n8n, and judgment-heavy steps belong in an AI step.

Linkeddit's side keeps its own clock. All five competitor monitors on the test project are set to weekly, and each shows a next run of 2026-11-04. So a weekly prompt always finds fresh data. n8n also documents an MCP Client Tool node that lets a workflow call tools on an external MCP server. This page does not rely on it, because it was not tested against Linkeddit.

Keep a data table of signal ids. The n8n connector lists create_data_table, add_data_table_rows and search_data_tables, and each Linkeddit signal has a stable id and a week_of date. Claude checks the table before adding a row, so an item already seen is not alerted again.

This also covers the first run problem the template store flags: the first pull is simply the first set of rows, not a seed that must be special cased. Keep alerts to high priority and review the watch tier weekly, so a busy week does not train people to ignore the channel.

Check three things. First, read the workflow Claude drafted: the Switch conditions, the alert recipients and the data table. Second, look at the test_workflow result on pinned data and confirm a low priority signal does not alert. Third, confirm which workflows n8n exposes to MCP and whether the connector asks permission before write tools run.

n8n documents result caps and a default five minute MCP execution timeout, and execute_workflow does not support multi-step forms or human approval steps. Keep the alert path short, and publish only after you have read it.