Products & Tools

Perplexity's Monitor API turns web changes into agent events

Perplexity launched Monitor on November 16, 2026, pricing web monitoring at $1 per 1,000 successful runs. Search and page checks now feed filtered events into applications and agents.

Perplexity Launches Monitor API to Automate Web Change Detection
Perplexity Launches Monitor API to Automate Web Change DetectionPexels / pixabay
By Rebecca Stone5 min read

Why it matters

  • Perplexity launched the Monitor API on November 16, 2026.
  • Web monitoring costs $1 per 1,000 successful runs; Page monitoring costs $1 per 1,000 URLs fetched.
  • A Page monitor tracks 1 to 10 known URLs, while checks can run every 15 minutes or on a custom cron schedule.
  • Run history is retained for 90 days, and webhook delivery is at least once.
  • Perplexity's eight-hour evaluation covered 60 topics and used a separate model to assess relevance and publication timing.

By Rebecca Stone

Perplexity launched the Monitor API on November 16, 2026, pricing web monitoring at $1 per 1,000 successful runs.

The service provides scheduled topic searches and webpage-change tracking. Filtered updates reach developers through the API or signed webhooks. Perplexity also makes Monitor available through its API Console.

Monitor combines recurring checks, saved state, duplicate removal and change detection in one developer service. Teams often assemble those pieces themselves to watch for account funding, breaking API changes or competitor pricing updates.

The release matters because it connects external web changes to downstream application or agent work. Developers can decide which developments merit action instead of polling sources and handling every alert separately.

Perplexity says: “Perplexity Decider filters updates through adjustable thresholds before any event reaches an application or agent.” That filter sits at the center of the product.

What does Monitor automate?

Perplexity lets developers create one monitor for repeated discovery or change detection. The service supports both automatic schedules and manual checks. Each run is one scheduled or on-demand check.

A Search monitor uses one query to find new web and news results about a topic, company or question.

A Page monitor watches 1 to 10 known URLs, including pricing pages, changelogs and policy pages. Perplexity does not allow the two modes to be combined in one monitor.

For a Page monitor, Perplexity saves an initial baseline and returns a text diff on later runs. That process works like version control for a small set of web pages: it isolates changed material instead of returning the full page.

Saved state between runs identifies new findings or page changes and deduplicates results across checks. This state lets recurring checks distinguish a fresh result from an item seen earlier.

Updates include a confidence score. Search updates also carry a title and snippet. Metadata can route an event to a particular user, workspace or agent.

How does Decider filter updates?

Perplexity describes Decider as a general-purpose classifier. It returns probability distributions over possible answers rather than generating free-form responses.

For Search findings, Decider checks:

  • Relevance
  • Source legitimacy
  • New information
  • Freshness

Each property has an independently adjustable threshold. An update is sent only when every applicable check clears its threshold. This lets developers set different standards for relevance, legitimacy, novelty and freshness instead of relying on one blanket filter.

For Page changes, Decider checks whether the change matters to the monitor’s optional plain-language goal. Developers can revise that goal without recreating the monitor. They can therefore change the definition of a useful change without resetting its saved state or schedule.

The announcement does not specify how Decider was trained or what its probability values mean at each threshold. It does establish that Decider operates before delivery and that thresholds, rather than a generated response, govern acceptance.

How do schedules and runs work?

Perplexity supports these schedules:

  • Every 15 minutes
  • Hourly
  • Daily
  • Weekly
  • Custom cron schedule

Creation makes the first run due immediately. Developers can request manual runs, pause a monitor and resume it later. These controls cover both rapid tests and long-lived recurring checks.

Perplexity lists web monitoring at $1 per 1,000 successful runs. Page monitoring costs $1 per 1,000 URLs fetched because one run can retrieve several pages. A run remains one scheduled or on-demand check regardless of how many pages it retrieves.

Monitor uses the same API key as the Search, Context and Agent APIs.

How are updates delivered?

Webhooks are optional because developers can read stored run results through the API. Signed webhooks automatically retry transient delivery failures.

Perplexity documents delivery as at least once. Events can arrive more than once and can arrive out of order. Receiving systems must deduplicate events and changes by their identifiers.

Monitor’s saved state removes duplicate findings across its checks. It does not change at-least-once delivery at the recipient, so developers still need deduplication logic there. The two layers cover different jobs: Monitor removes duplicate results across checks, while the recipient removes repeated deliveries.

Run history is retained for 90 days. Deleting a monitor removes its run history and saved state while stopping future scheduling. Repeated source or webhook failures, insufficient credits or organizational policy can trigger an automatic pause.

Organizations requiring zero data retention cannot create, resume or run monitors. Perplexity calls the exclusion a material enterprise restriction. The policy blocks the entire monitor lifecycle for those organizations, including initial setup and manual execution.

What did the evaluation show?

Perplexity reported that Monitor delivered “the most fresh, relevant results across the widest topic range” in an eight-hour evaluation against other Search API providers. The company said many results did not appear from any other tested provider.

The evaluation covered 60 topics drawn from production Search API queries, including company news, technology releases and regulation. A separate model assessed relevance and publication timing.

Perplexity’s announcement does not name the competing providers, publish comparative scores or identify the separate model. It also does not report precision, latency or false-positive rates. The result supports Perplexity’s claim of breadth across its test set, but it remains a company-run evaluation rather than an independent comparison.

What can Monitor trigger?

Perplexity describes production uses that connect detection to action:

  • Funding announcements, executive hires or product launches can trigger account research, CRM updates and outreach through connected go-to-market agents.
  • Equity research monitors can follow earnings releases, guidance changes, acquisitions and adverse media.
  • Regulatory workflows can summarize and route changes on regulator websites to compliance teams.

These examples span revenue, investment and compliance. In each case, Monitor identifies an external change and sends a filtered event into another system. It serves as a trigger layer, not the complete business workflow.

The company frames Monitor as a persistent input stream for agents that need to act on external change. Its enterprise significance lies in joining filtered detection to workflow execution, rather than merely exposing another search call.

Saved-state architecture, at-least-once webhook behavior and the exclusion of zero-data-retention organizations define the practical boundaries of deployment.

For Perplexity, the next proof point is operational: enterprises need filtered events to trigger reliable downstream work despite duplicate and out-of-order deliveries, 90-day history and strict data-handling exclusions.

Source: unite.ai

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

728 articles

Related articles

  1. Perplexity launches Teammates: AI agents that own ongoing work, not just chats
  2. AI 'bossware' is rating teachers' work — and stressing them out
  3. Goodfire's inside-out AI monitors cost $185 per million checks
  4. Perplexity Launches Multiagent Mode to Coordinate Frontier AI Models
  5. Perplexity’s Orchestrator retries failed requests across providers

« Previous articleNext article »