Perplexity Monitor API watches pages for useful changes
Perplexity’s 16 November announcement introduces the Monitor API: scheduled checks that send relevant web updates to an application through webhooks. For a commerce team, the useful starting point is…
Perplexity’s 16 November announcement introduces the Monitor API: scheduled checks that send relevant web updates to an application through webhooks. For a commerce team, the useful starting point is a small watchlist of pricing, shipping or supplier-policy pages.
What the service handles
According to Perplexity’s product description, developers can monitor a topic across the web or track up to ten specified pages. Page monitoring saves an initial baseline and returns later text differences. A plain-language goal helps filter which changes matter.
Checks can run every 15 minutes, hourly, daily, weekly or on a custom cron schedule. The service deduplicates results and sends signed webhooks, with retries for transient delivery failures. Its relevance thresholds are adjustable, and updates include confidence scores.
Count pages when estimating cost
Web monitors cost $1 per 1,000 successful runs. Page monitors cost $1 per 1,000 URLs fetched: one run can fetch several pages. The API uses the same key as Perplexity’s Search, Context and Agent services.
As a simple planning example, ten pages checked once an hour produce 240 URL fetches a day. At the announced rate, that is $0.24 a day in page-fetch charges, before any separate processing or infrastructure costs. This is arithmetic from the published unit price, not a measured bill.
Keep the decision with the merchant
Our suggested pilot is to send each alert into a review queue with the original URL, observed change and timestamp. Compare the changed page with the relevant product, market and delivery terms before taking action. A promotion on one variant should not trigger a catalogue-wide price reduction.
Test the watchlist against changes you already know about, and record both missed changes and noisy alerts. A confidence score is not a guarantee that the source is correct or that a commercial response is appropriate.
For context, see Aravind Srinivas’s roundup of the Monitor, Orchestrator and Context APIs. These are separate products; this article concerns scheduled monitoring.
More from AI
See all →
