A familiar situation: once a week, a manager manually opens five competitors’ websites and checks prices against a spreadsheet. Three or four days later, they finally notice someone launched a sale. By then, some customers have already moved to the cheaper option. Checking prices manually every day simply isn’t practical once your catalog runs into the hundreds of items.
That’s exactly the problem an AI agent built in n8n solves. It visits competitor websites on a schedule, pulls the current prices, compares them to the previous snapshot, and sends an alert the moment something changes. The agent doesn’t just scrape raw text, either — it uses an AI model to parse inconsistent page layouts and turn messy HTML into clean, structured price data.
This article breaks down how such an agent is built, what you need to run it, what it costs, and the mistakes people most often make on their first attempt.
What’s inside
- How an AI agent for price monitoring works in n8n — a step-by-step breakdown of nodes, from trigger to alert
- What you need to implement it: access, APIs, data storage
- What setup and monthly maintenance cost
- Why n8n beats ready-made SaaS price trackers as the number of competitors grows
- Common mistakes on the first build, and how to avoid them
- When it’s worth handing price monitoring to a specialist
How the agent is built, technically
The workflow is made up of six logical blocks, each handling one narrow job.
Schedule Trigger. The agent runs on a schedule — twice a day, for example, morning and evening. For competitors who change prices often, you can shorten the interval to a few hours, but it’s important to keep request frequency to someone else’s site within a reasonable range.
HTTP Request to competitor pages. This node pulls the HTML of the relevant product pages. For sites with bot protection or heavy JavaScript rendering, teams typically route this through a third-party scraping API using the same HTTP Request node — simpler than running a headless browser inside n8n.
AI processing (an OpenAI/Claude node, or an HTTP Request to the model’s API). Raw HTML looks different on every competitor’s site: sometimes the price sits in a span, sometimes in a data attribute, sometimes it’s hidden behind a discount. Instead of writing a custom parser for every site, the page text is passed to the AI model with instructions to extract the product name, current price, and stock status. That removes the need to constantly rewrite selectors every time a competitor redesigns their site.
IF/Switch — comparing against the previous value. This node compares the newly fetched price against the one stored from the last run. If the difference is below a set threshold (say, 1-2%), the workflow ends without sending a notification — filtering out noise from rounding and short-lived glitches on the competitor’s site.
Writing to storage (Google Sheets/Airtable/Postgres). Every price point is saved with a date and source, building a history that shows trends over time rather than just a single snapshot.
Action — sending the alert. When a meaningful price change is detected, a Telegram, Slack, or email node sends a short notification to the responsible person: which product, which competitor, how much the price moved, and what your current margin looks like.
Error Trigger. A separate workflow catches failures in the main one: the competitor’s site is temporarily down, its structure changed, or it blocked the request. Without this piece, the agent can quietly return empty data for months while you assume prices simply haven’t moved.
What you need to implement it
To get this running you’ll need n8n itself — self-hosted on a VPS or n8n Cloud — access to competitor pages (either direct requests or a scraping API for protected sites), an AI model API for parsing the data, and storage for the price history, whether a spreadsheet or a database. For alerts, the Telegram Bot API or email is usually enough.
It’s worth noting the legal side too: monitoring publicly available prices is generally acceptable, but it’s worth checking each site’s terms of service and avoiding excessive load from overly frequent requests.
What setup and maintenance cost
| Cost item | Approximate cost |
|---|---|
| VPS for self-hosted n8n | $5-10/month |
| n8n Cloud (alternative to a VPS) | from $20/month |
| Scraping API (for protected sites) | $30-50/month |
| AI model API (data parsing) | $5-15/month |
| Total | ~$40-75/month |
The key point: this figure barely moves as you add more products or competitors within a reasonable range — unlike ready-made SaaS price trackers, where pricing is usually tied to the number of SKUs monitored and climbs as your catalog grows.
Why n8n instead of a ready-made builder
Ready-made no-code platforms like Make.com charge by the number of operations. Checking 20 products across 5 competitors twice a day quickly adds up to thousands of operations a month and a noticeable bill. Running n8n on your own server, by contrast, is a fixed cost regardless of how many checks you run. A self-hosted setup also gives you full control over the scraping logic and the AI prompts, so you can fine-tune them to your specific niche instead of working around someone else’s builder’s limitations.
Common mistakes
The first mistake is trying to cover fifty competitors and an entire catalog from day one. It’s smarter to start with 3-5 key competitors and your highest-margin products, then expand once the workflow has run reliably for at least a month.
The second is skipping the Error Trigger. Without it, a parsing failure on one of the sites can go unnoticed for weeks.
The third is skipping the test period — launching the workflow against the full page list instead of first checking it against 2-3 product pages to confirm the AI is extracting prices correctly.
The fourth is ignoring request-rate limits, which gets the server’s IP address banned by the competitor. The fifth is keeping data only inside the workflow itself rather than in a separate table — so the history simply disappears on the next restart.
When it’s worth handing the build to a specialist
If competitors run serious anti-scraping defenses — captchas, geo-blocking, dynamically generated JavaScript content — setting this up yourself will take disproportionately long. The same applies when you need to track hundreds of SKUs across dozens of competitors at once, integrate the data into an internal ERP or CRM, or build automatic price adjustments on top of the monitoring itself — the cost of a mistake is high there, and it’s better to bring in an n8n specialist.
A practical example
A consumer electronics store with 200 SKUs was tracking prices for five key competitors by hand: a manager spent about 6 hours a week on it, and reacting to changes took 3-5 days. During that lag, the store lost an average of 2-3% margin on items that had quietly become uncompetitive.
After rolling out the n8n agent, checks run twice a day, and a Telegram alert fires whenever a price shifts by more than 3%. Reaction time dropped to about an hour, the manager’s time was almost entirely freed up — roughly 24 hours a month — and running the workflow costs around $50 a month.
Conclusion
An AI agent for competitor price monitoring in n8n handles a job that’s genuinely impractical to do by hand: continuously watching for changes across dozens of pages. The cost stays fixed rather than growing with your catalog, and implementation takes days rather than months — as long as you start with a small set of competitors and build in an Error Trigger from day one.
If you want to go deeper
- AI Agent: Website Analysis, Weekly Report
- Small Business Automation: Where to Start
- AI Agent for Online Stores
FAQ
Is it legal to scrape competitor prices? Monitoring publicly available prices generally doesn’t break the law, but it’s worth checking the specific site’s terms of service and avoiding excessive load on its servers.
Do I need a developer to set this up? A basic version covering 3-5 competitors without heavy anti-scraping defenses can be built yourself, following the workflow structure described here. For protected sites and larger scale, a specialist is worth bringing in.
How does the agent handle sites that block scraping? Usually through a third-party scraping API that emulates a real browser and gets past basic protections. Complex cases involving captchas may need additional setup.
Can this monitor prices on marketplaces like Amazon or eBay? Technically yes, but marketplaces often have stricter protection against automated requests, so a workflow like this usually needs a scraping API and more careful request pacing.
What if a competitor changes prices several times a day? Just increase the Schedule Trigger’s frequency — checking every 2-3 hours, for example — while staying within reasonable request limits.
How long does it take to set up the agent from scratch? For 3-5 competitors without heavy anti-scraping defenses, usually 1-3 days, including a test period on 2-3 product pages before the full rollout.
Leave a Reply