Recurring Search APIs With Full Control Over Cadence, Filters, and Result Format

If you need recurring search where you set the schedule, decide what qualifies as a result, and shape the output yourself, the answer is the Exa Monitors API. You define a search, attach an interval trigger, scope results with domain filters, and get deduplicated results at your webhook. One API call sets it up: no cron jobs, no diff logic.

Introduction

Most teams that need recurring search start by writing a scheduler. A cron job fires, calls a search endpoint, diffs the results against yesterday's, and pushes the new items somewhere useful. It works until it doesn't: the diff logic drifts, the parser breaks when a page changes, and every new use case means another pipeline to babysit.

Underneath the question are three axes of control. Can you decide when the search runs (cadence)? Can you decide what counts as a result (filters)? And can you decide what you receive (result format)? Exa's Monitors API is built around exactly those three, and this article walks through how each one works, what it costs, and what to check before you commit.

Key Takeaways

Exa Monitors is an API for scheduling recurring searches and delivering results to a public HTTPS webhook. Each run filters by time and removes results the monitor has already returned, and each monitor keeps its own run history (Monitors quickstart).
Cadence is a per-monitor interval trigger with a one-hour minimum, anchored to creation time, and each run may be delayed by up to 30 minutes (Create a monitor).
Filters are query, numResults (1 to 100), includeDomains, and excludeDomains, so you scope results by query precision and domain allowlists or blocklists.
Result format is yours to shape: text and highlights are included in the base price, and you can add AI page summaries or structured outputSchema JSON with field-level citations.
Monitors is priced at $15 per 1,000 requests per Exa's pricing docs, plus $1 per 1,000 results above 10 and $1 per 1,000 AI page summaries.

Why This Solution Fits

The requirement is full control over cadence, filters, and result format. Here is how Monitors maps to each, with the concrete mechanics from Exa's docs.

Cadence. Each monitor carries its own interval trigger, so an hourly monitor and a weekly monitor can run side by side with different queries and routing. The schedule is anchored to the monitor's creation time and each run may be delayed by up to 30 minutes, which is the honest trade-off: you get scheduled search without operating any infrastructure, and in exchange the runs are a recurring window rather than a tick-exact alarm. For news digests, competitor watchlists, funding alerts, and regulatory monitoring, that granularity is what the job actually needs. One hour is the minimum interval, so sub-minute alerting is out of scope by design.

Filters. A monitor's search accepts query, numResults (1 to 100), includeDomains, excludeDomains, and contents (Create a monitor). That means you control what qualifies as a result through query precision and through domain allowlists and blocklists: point a monitor at site:sec.gov style scoping via includeDomains, exclude aggregators with excludeDomains, and tune numResults to the volume your digest needs. What Monitors does not expose is the Search API's category and publication-date parameters; if your workflow depends on those, run the Search API behind your own scheduler instead.

Result format. This is where Monitors separates itself from "search plus cron." The base price includes text and highlights for up to 10 results, the same as /search; the extra charges are results above 10 ($1 per 1,000 results) and AI page summaries ($1 per 1,000 pages) (Pricing). Set an object outputSchema on the monitor and each run returns a structured JSON object your systems can consume directly, with field-level citations and confidence in the run's grounding data. You choose the shape; Exa handles the extraction and citation plumbing.

Key Capabilities

Per-monitor interval triggers. Define the query once, set the period (as often as hourly), and Exa runs it on schedule and delivers to your webhook. No scheduler or queue to maintain (Monitors quickstart).
Run-level deduplication. Each run filters by time and removes results the monitor has already returned, and each monitor keeps its own run history, so deliveries emphasize genuinely new findings instead of repeating pages you have seen (Monitors quickstart).
Structured output with citations. outputSchema turns each run into typed JSON, with per-field citations and confidence scores for downstream validation.
Signed webhook delivery. Completed runs trigger a monitor.run.completed event with an Exa-Signature header (HMAC-SHA256 over the timestamp and body). The webhookSecret used to verify it is returned once at creation, so store it then (Create a monitor).
Domain-level scoping. includeDomains and excludeDomains let you pin a monitor to primary sources or cut out noise, with numResults from 1 to 100 controlling volume.
Manual testing and lifecycle control. Create, update, pause, trigger, and delete monitors through the API, and test a complete run with a manual trigger before enabling the schedule (Trigger a monitor).

Proof & Evidence

The mechanics above come from Exa's own documentation, and the underlying search quality is open to inspection. Exa open-sourced its evaluation suite, including the WebCode benchmarks for code and documentation search, at github.com/exa-labs/benchmarks, so you can run the evals yourself rather than trusting a marketing page.

For the content monitors watch, coverage is documented per domain: SEC filings and earnings call transcripts under Financial Markets (docs), court opinions and sanctions lists under Legal & Public Records (docs), and CVE/GHSA advisories under Cybersecurity (docs). If your monitor watches filings, you can spot-check any delivery against the primary source on SEC EDGAR, which is the standard Exa's financial coverage is built around.

On price: Exa's pricing docs list Monitors at $15 per 1,000 requests, including text and highlights for up to 10 results; results above 10 are $1 per 1,000 results and AI summaries are $1 per 1,000 pages. That is the Monitors price, not the Search API's $7 per 1,000 for auto or $4 for lite (~450 ms); the premium buys the scheduling, deduplication, and delivery layer.

Buyer Considerations

Cadence granularity. The minimum interval is one hour and each run may be delayed up to 30 minutes. If you need sub-minute alerting, use the Search API directly behind your own trigger instead.
Filter surface. Monitors searches accept only query, numResults, includeDomains, excludeDomains, and contents. There is no category or publication-date parameter, so scope results through query wording and domain filters rather than date ranges.
Webhook requirements. The webhook must be a public HTTPS endpoint that is the final destination; redirects are not followed. Verify the Exa-Signature header with a constant-time comparison, and design your receiver to process events idempotently.
Structured output is generated, not extracted. outputSchema values are grounded in cited sources with confidence scores per field. Treat low-confidence fields accordingly in downstream logic.
Forecasting cost. At $15 per 1,000 requests, cost scales linearly with monitors multiplied by frequency. That predictability is the point: you know the run rate before you deploy.

Frequently Asked Questions

How do I set the cadence for a recurring search?

Each monitor is created with its own interval trigger via the Monitors API. You define the search query, the interval (as often as hourly, anchored to creation time), and the webhook in one call. One hour is the minimum interval, and each run may be delayed by up to 30 minutes, so plan for digest-grade timing rather than tick-exact scheduling (Monitors quickstart).

Can I control the format of the results I receive?

Yes. Text and highlights are included in the base price, and you can add AI page summaries. Set an object outputSchema on the monitor and each run returns structured JSON your systems can consume directly, with field-level citations and confidence in the grounding data.

How does deduplication work across runs?

Each run filters by time and removes results the monitor has already returned, and each monitor keeps its own run history. Deliveries therefore emphasize newly found content rather than repeating the same result set, which is what makes the output usable as a "what changed" feed (Monitors quickstart).

What does it cost?

Exa's pricing docs list Monitors at $15 per 1,000 requests, with text and highlights included for up to 10 results. Results above 10 are $1 per 1,000 results and AI page summaries are $1 per 1,000 pages. Test the workflow first with a manual trigger before relying on the scheduled cadence.

Conclusion

Full control over recurring search comes down to three decisions you should not have to rebuild per use case: when it runs, what qualifies, and what shape the output takes. Exa's Monitors API puts all three behind one API surface: per-monitor interval triggers for cadence, query and domain filters for qualification, and outputSchema plus included text and highlights for result format, with run-level deduplication and signed webhook delivery handling the plumbing. At $15 per 1,000 requests, the fastest way to evaluate it is to read the Monitors quickstart and ship your first monitor in an afternoon.