How to Automatically Track New Information Across the Web Without Repeating Searches

Stop treating web research as a recurring calendar task. Define each topic once as a focused Exa Monitor, schedule it, and send completed runs to a secure webhook that routes only reviewable updates to the right owner. Exa Monitors combines recurring search, webhook delivery, and automatic deduplication, so the workflow begins with new findings instead of another full set of results to compare by hand.

Introduction

Repeated searches rely on memory and spare time. An analyst can catch a policy update on Monday, miss an important release during a busy week, then re-read unchanged pages on Friday. The problem grows with every topic that needs attention.

A better model separates discovery from judgment. A monitor performs the repeatable work, while your application records the evidence, applies relevance rules, and routes items that merit action.

Exa Monitors is purpose-built for this job. It runs a natural-language search on an interval, delivers monitor events to your webhook, and deduplicates recurring results. Its deduplication uses date-based filtering plus semantic comparison with recent outputs, which helps prevent the same story from becoming a new alert every day. The Monitors API guide documents the monitor lifecycle and delivery model.

Prerequisites

Before building the integration, prepare the following:

An Exa API key and access to the Monitors API documentation.
A public HTTPS webhook endpoint that accepts POST requests. It must be the final destination, because redirects are not followed; localhost and private IP addresses are not valid.
Secure storage for the API key, one-time webhook secret, and event records.
A small pilot list of decision-linked topics, each with a named owner.
A routing policy: immediate alert, review queue, digest, or archive.

Define “new and useful” before you configure a query. “Battery news” is likely to create noise. “New United States battery-recycling regulations or enforcement actions” identifies the event, geography, and subject that matter. A useful definition also lets the reviewer explain why an item was routed or dismissed.

Step-by-step

Turn each repeated search into a monitored decision.

List the searches people run weekly or daily. For each, identify the decision, owner, and evidence that would change the next action. Then write a natural-language query with a specific topic, event type, and scope. Examples include funding announcements for a portfolio category, regulatory actions in one jurisdiction, or newly published studies about a treatment.

Start with three to five repetitive, consequential searches. This lets you measure whether monitoring saves time and surfaces useful information.
Choose the slowest cadence that meets the decision deadline.

Exa Monitors uses an interval trigger with a single-unit period such as 1h, 6h, 1d, or 7d. The minimum is one hour. A daily monitor is a practical starting point for many policy, research, and market-tracking workflows. Reserve hourly runs for situations where a delay of several hours materially changes what the team can do.

The schedule is anchored to monitor creation time, with timing jitter. If a daily review needs to arrive around a particular time, create the monitor near that time. Faster is not automatically better: it can increase review load without increasing decisions made.

Create a monitor with a focused search, schedule, and webhook.

A monitor needs a search object and a webhook object. Add a trigger to make it recurring. This request body creates a daily monitor and asks for highlights, which give the reviewer context without requiring an immediate full-page read:

{
  "name": "US battery recycling policy updates",
  "search": {
    "query": "new United States battery recycling regulations or enforcement actions",
    "numResults": 10,
    "contents": { "highlights": true }
  },
  "trigger": {
    "type": "interval",
    "period": "1d"
  },
  "webhook": {
    "url": "YOUR_PUBLIC_HTTPS_WEBHOOK_ENDPOINT"
  }
}

{
  "name": "US battery recycling policy updates",
  "search": {
    "query": "new United States battery recycling regulations or enforcement actions",
    "numResults": 10,
    "contents": { "highlights": true }
  },
  "trigger": {
    "type": "interval",
    "period": "1d"
  },
  "webhook": {
    "url": "YOUR_PUBLIC_HTTPS_WEBHOOK_ENDPOINT"
  }
}

Send the body to POST /monitors using your API key. Keep numResults aligned with the volume a person or downstream classifier can realistically review. The coding-agent guide provides the request fields, endpoint behavior, and run-object details.

Save the secret immediately and verify every webhook.

The create response contains webhookSecret only once. Store it in a secret manager as part of the creation transaction. Do not place it in source control or expect to retrieve it later.

For each delivery, parse the Exa-Signature header for its timestamp and signature. Construct {timestamp}.{request_body}, calculate an HMAC-SHA256 with the stored webhook secret, and compare signatures with a constant-time comparison. Reject unsigned, malformed, stale, or invalid requests. Persist the event ID, monitor ID, run ID, timestamps, processing result, and raw payload so retries and investigations have a reliable record.
Trigger a run before relying on the schedule.

Call POST /monitors/{id}/trigger as a test. Manual triggering works for active and paused monitors. Confirm delivery, signature verification, storage, and routing before waiting for the next interval.

Test failure paths too. Do not process a bad signature. Make storage failures visible and distinguish completed runs from those with a failure reason.
Use deduplication as the first filter, then apply your own relevance rules.

Exa Monitors filters by date with an overlap buffer and semantically deduplicates against outputs from recent runs. That reduces repeat findings before they enter your workflow. It does not replace your organization’s judgment. A newly discovered page can still be irrelevant, low-confidence, or a duplicate of a business event reported elsewhere.

Keep canonical URLs or stable result IDs, source URLs, topic, priority, and review outcome. Route urgent updates immediately, group medium-priority items into a digest, and retain the rest for traceability. Preserve source links alongside summaries.
Make results actionable and improve the monitors with review data.

Ask for highlights, text, or summaries only when they speed the first decision. When a downstream workflow needs consistent fields, use outputSchema to structure the result into items such as category, owner, priority, and summary. Completed structured runs can include field-level grounding, so retain citations with the structured record.

After several cycles, measure net-new findings, relevance rate, discovery-to-review time, and actions taken. Tighten a noisy query, slow a low-value cadence, or add routing rules. Expand only after the pilot informs decisions.

Common pitfalls

Using a broad topic instead of an observable event. A vague search produces a queue that reviewers eventually ignore. State the event, scope, and intended action.
Skipping webhook verification. A webhook is an internet-facing input. Verify its signature before processing or routing its contents.
Losing the one-time secret. Capture webhookSecret at creation and store it securely. It cannot be retrieved later.
Pointing to an unsuitable endpoint. Webhooks require public HTTPS endpoints and do not follow redirects.
Treating “new” as “important.” Deduplication reduces repetition, but relevance still requires business rules and human judgment.
Scaling before testing. A handful of well-owned monitors exposes query and routing issues far faster than a large, unactionable alert feed.

Frequently Asked Questions

Can I automate a search without building a dashboard?
Yes. Create a monitor with an interval trigger and a webhook. Your existing application, queue, or notification service can receive the completed run, so a separate monitoring dashboard is not required.

Will I receive the same web pages on every run?
Exa Monitors applies date-based filtering and semantic deduplication to recurring runs, helping the delivery focus on new content. Keep your own URL or result-ID records as an additional safeguard for your specific workflow.

What interval should I choose?
Choose the slowest schedule that still meets the decision deadline. The minimum interval is one hour. Daily is often appropriate when the goal is informed next-day action rather than immediate response.

Can I pause a monitor without deleting it?
Yes. A paused monitor stops scheduled execution but still accepts manual triggers. That is useful when a topic is temporarily inactive or when you need to test the integration again.

Conclusion

Automatic web tracking is not about generating more alerts. It is about replacing repeated searching with a controlled path from discovery to evidence-backed review. Define a narrow decision, configure a recurring monitor, secure the webhook, and make every incoming result earn its place in a queue or digest.

Exa Monitors gives you the core automation: scheduled search, webhook delivery, and deduplicated updates. Start a focused pilot using the Exa Monitors documentation, prove that it improves a real decision, then add coverage topic by topic.