What happened
In a post dated October 29, 2026, Exa describes Monitors as a way to specify what to watch on the web and have the service deliver new findings to an agent at a chosen cadence. The product is positioned against two familiar approaches: email-oriented alerts such as Google Alerts, which Exa says lack an API or webhook for automation, and agents that periodically rerun searches as cron jobs. In the latter arrangement, the company argues, searches often return nothing new while still incurring search and token costs.
Exa says Monitors searches its index, which it describes as covering 211 million businesses, 3.2 million podcast episodes, 1.27 billion product listings and more than 130 billion web pages. Each run has three stages. Retrieve performs a fast search across the index with a time window matched to the monitor’s cadence. Judge drops pages already seen by that monitor, then checks each remaining page for relevance, source legitimacy, new information and event timing. Compose turns pages that pass into either a short summary or JSON matching the customer’s outputSchema; the output includes sources and a confidence level for each field. Exa says results go to a webhook, so an agent does not need to poll.
The principal commercial change in v2 is price: Exa says monitors now cost $2 per 1,000 runs, down from $15, a reduction of more than sevenfold. It says existing implementations do not need to change. Its example uses the Python Exa client to create a monitor with a query, a daily interval and a webhook subscribed to monitor.run.completed; each completed run sends new pages, a summary or structured output, and citations. The post also suggests adding metadata to route an update to the right user or workspace when one agent serves many people. It points readers to the Monitors quickstart or an API key to begin.
Exa lists monitoring product prices and stock, news topics, competitor activity and people’s career moves as possible uses. It says podcast transcripts can surface a competitor mention that appears deep in an episode, and gives examples including prediction-market evidence collection, portfolio-company partnership tracking, sales triggers and adverse-media screening. These are use cases described by the vendor, not independently documented customer results. Exa reports that a side-by-side run of v1 and v2 on real production monitors delivered only relevant updates in v2 and made monitors “noticeably quieter,” but provides no sample size or measurement method.
What the archive adds
An earlier Telegram post on April 3 described Exa Monitors as scheduled search that sends new results to a webhook. This is a new version and pricing fact on that known story, not a wholly new product thread. Our related Exa Agent Ultra note concerns exhaustive list-building in agent search; it is adjacent context, not evidence about Monitors v2.
Trend signal
The update supports agent economics moves to completed work only weakly: Exa’s price is per run, not per verified outcome, but its product pitch is to spend less on empty searches and deliver useful changes rather than repeated activity. The post does not show whether customers’ cost per completed monitoring task falls in practice.
Why now
Exa frames v2 as a response to the cost and noise of agents rerunning searches on a schedule, and to the limits of alerts designed for inboxes rather than automation. The post supplies no further explanation for the timing of the release.
Connections to other events
The earlier April Telegram post described Monitors’ basic scheduled-search-to-webhook model. The v2 announcement retains that interface while emphasizing deduplication and a lower run price. Exa’s separate Agent Ultra note also treats search as a component of agent workloads, but does not establish a technical dependency between the products.
Why it matters
For a team building a monitoring agent, the practical choice is not only which search API to call, but where to handle state, deduplication and delivery. If Exa’s filters perform as described, a webhook feed that remembers prior results can reduce scheduled empty searches and avoid sending an agent the same coverage repeatedly. The price reduction makes experimentation with more monitors less expensive on the vendor’s stated run-price basis.
The trade-off is that teams delegate part of their signal-quality logic to the provider. The four checks—relevance, source legitimacy, novelty and event timing—sound useful, but the announcement does not expose their thresholds or guarantees. Teams should test representative queries against their own definition of a meaningful change, particularly for compliance or sales workflows where a missed event or false alert has a real cost. Structured output with citations and confidence fields may help downstream review, but it does not itself verify correctness.
Limits and uncertainty
The announcement is a vendor account. It does not provide benchmark methodology, evaluation sets, sample sizes, false-positive or false-negative rates, or a quantified comparison of v1 and v2. “Every update” being relevant refers to Exa’s side-by-side production test as described in the post, not a general guarantee. The supplied material also does not specify whether the $2 price includes all search, webhook, storage or downstream model costs, nor does it state monitor limits, retention behavior, service-level commitments or regional availability. Teams should confirm pricing and operational limits in current product documentation before relying on the figures.
Skill to build now
Build a small replay test before routing monitor events into production actions. Save a representative set of source pages and known updates, run queries at the cadence you expect to use, then review which events are emitted, which are suppressed and whether citations support the JSON fields. Record missed changes and duplicate alerts, and keep consequential actions behind human review until the test shows the feed matches your team’s threshold for a real signal.
Reader takeaway
Monitors v2 is an API-level monitoring update, not simply another alert inbox: it keeps state, filters results and pushes structured changes to agents. The price falls sharply per run, but Exa’s claims about quieter, relevant output remain vendor-reported. Test the quality of the changes you actually need, not just the cost of each run.