The Cost of Building Financial Research Infrastructure With a Web-Data API

Building financial research infrastructure on a web-data API costs roughly $70 to $300 a month for a team running about 10,000 standard searches and a few hundred agent runs, priced per request rather than per seat. That covers 10,000 auto searches at $7 per 1,000, deep searches at $12 per 1,000, and fixed-effort Agent runs from $0.012 each.

Introduction

The traditional way to equip a research team is a terminal subscription: a fixed, per-seat annual fee that bundles data, a screen, and some scripting. The alternative is to assemble your own stack on top of a web-data API and pay only for what you use. The question most teams ask first is simple: what does that actually cost?

The honest answer is that API pricing is usage-based, so it is forecastable only if you know your query volume. This article walks through the published prices for the Exa surfaces a financial research workflow typically needs, shows how the bill scales with volume, and flags the add-on fees that are easy to miss.

Key Takeaways

Exa's Search API costs $7 per 1,000 requests on the default auto type (about 1 second typical latency in the docs), and $12 per 1,000 for both deep-lite and deep (pricing).
The lite type is the low-latency option at $4 per 1,000 requests and roughly 450 ms, with cached content only and no AI page summaries (Search quickstart).
On /search, text and highlights are included in the base price for up to 10 results; the only extra charges are results above 10 ($1 per 1,000 results) and AI page summaries ($1 per 1,000 pages).
The Agent API, built for multi-step research like screens and company lists, prices by effort level: fixed-effort runs cost $0.012 (minimal) to $1.00 (xhigh) per request, while the auto default and ultra are metered at $0.10 per Agent Compute Unit plus $0.005 per search, with default per-run caps of $5 and $20.
Exa's free tier includes $10 in credits every month (up to 2,500 Lite searches), plus a one-time $10 onboarding bonus, with no payment method required, so you can prototype before spending anything.

Why This Solution Fits

A terminal subscription bundles data access, display, and tooling into one seat price. A web-data API unbundles the data layer and lets you build exactly the research surface your team needs: a screening script, a watchlist pipeline, an earnings-prep workflow. The cost profile follows the work, not the headcount.

Exa's financial data coverage is documented, not implied: the Financial Markets data page lists SEC filings and reported financials, earnings call transcripts with speaker attribution, prices, analyst estimates, funding announcements, and economic releases. Ordinary financial queries reach that broader coverage, while category: "financial report" narrows results to filings and reports. For filings and primary documents, the underlying source is public: SEC EDGAR full-text search covers filings submitted electronically since 2001, so a web-data layer that indexes filings is working over a checkable public record rather than a proprietary black box.

The pricing model also matches how research work actually happens. Most days a team runs a few hundred queries; earnings season or a deal sprint might run tens of thousands. Per-request pricing means the quiet months cost little and the busy months scale linearly.

Key Capabilities

Two Exa surfaces cover most of a financial research build, each with a specific role:

Search API is the workhorse for lookups and grounding: filings, transcripts, funding news, and estimates. The auto type returns in about 1 second typical at $7 per 1,000 requests; when a question needs a synthesized answer on top of the ranked results, deep with outputSchema runs 4 to 15 seconds at $12 per 1,000 (the ranked pages stay in results). For high-volume, latency-sensitive work like autocomplete in an internal tool, lite returns in roughly 450 ms at $4 per 1,000, though it serves cached content only and does not support outputSchema. Structured output via outputSchema on auto returns generated values in output.content with field-level citations and confidence in output.grounding, capped at 10 properties and 2 levels of nesting, and adds a synthesis step to the request.
Agent API handles the multi-step jobs a screen or account list needs: multiple searches, cross-referencing, and structured enrichment in one async run. Fixed effort levels price at $0.012 (minimal), $0.025 (low), $0.10 (medium), $0.50 (high), and $1.00 (xhigh) per request; the auto default and ultra meter at $0.10 per Agent Compute Unit plus $0.005 per search, capped at $5 and $20 per run by default. Exa's pricing estimator on its GTM use case page shows $0.10 per account at medium effort, before contact enrichment ($0.02 per email, $0.07 per phone) and Exa Connect provider fees.

Proof & Evidence

The clearest cost evidence is Exa's published price list. Per the pricing documentation, Search requests cost $4 per 1,000 on lite, $7 per 1,000 on auto, and $12 per 1,000 on deep-lite and deep; Contents pages are $1 per 1,000 per content type. The free tier includes $10 in monthly credits (up to 2,500 Lite searches) plus a one-time $10 onboarding bonus, with no payment method required, which lets you measure your real query volume before spending anything.

Latency is published too, which matters when you are sizing an interactive tool. Exa's Search quickstart lists lite at about 450 ms and auto at about 1 second typical, and Exa's SealQA comparison benchmark reports auto at 1,502 ms (about 1.5 seconds) median (Exa vs Parallel, updated October 19, 2026); neither figure is a response-time guarantee, and the pricing page summarizes the range as "Configurable latency: 450ms to 1s". A screening tool that feels instant is feasible on the standard tiers, and the per-request price stays the same whether the tool serves one analyst or twenty.

Buyer Considerations

Estimate volume first. At 10,000 auto searches a month, Search costs about $70. The same volume on deep costs about $120, and on lite about $40. Your architecture choice (link lists versus synthesized answers) moves the bill more than anything else.
Budget the Agent add-ons. Any Agent cost estimate should say "before contact enrichment and Exa Connect provider fees." Contact enrichment adds $0.02 per email and $0.07 per phone; Exa Connect partner calls (including Financial Datasets for SEC filings and financials) are billed on top of the run. Throughput also depends on your account's Agent limits, so benchmark before estimating concurrency.
Use budgets with auto or ultra effort only. Exa's docs state that budget.maxCostDollars ($1 to $100) works only with effort: "auto" or effort: "ultra", and budget.maxDurationSeconds (300 to 10,800 seconds) works only with ultra. Fixed efforts do not accept a budget parameter.
Mind the schema caveats. Agent structured fields can be null even when marked required, and citations appear in output.grounding when emitted, not on every data point. Build validation into your pipeline rather than trusting shape alone.

Frequently Asked Questions

How does per-request API pricing compare to a per-seat terminal subscription?

A terminal subscription charges a fixed annual amount per user regardless of usage. Exa's API charges per request: $4 to $7 per 1,000 standard searches (lite or auto), $12 per 1,000 deep searches, and $0.012 to $1.00 per fixed-effort Agent run, with the default auto mode and ultra usage-metered (default per-run caps of $5 and $20). A small team with moderate volume often pays less with the API; a large team running heavy interactive sessions all day should model both before switching.

What is the cheapest way to prototype a financial research tool?

Start on the free tier, which includes $10 in credits every month (up to 2,500 Lite searches) plus a one-time $10 onboarding bonus, with no payment method required. That is enough to test query patterns, tune outputSchema shapes on auto, and measure real latency before any spend.

Do I need the Agent API, or is Search enough?

Search is enough for lookups and single-question grounding, including synthesized answers via the deep-lite and deep types with outputSchema. Use the Agent API when a task needs multiple searches, cross-referencing, and structured enrichment, such as screening a company list or building comps. Exa's own Deep Search documentation recommends Agent for long-running research and multi-hop enrichment rather than pushing a deep search type past its design.

What hidden costs should I watch for?

Three: results above 10 per Search request ($1 per 1,000 results), AI page summaries ($1 per 1,000 pages, not available on lite), and Agent add-ons (contact enrichment at $0.02 per email and $0.07 per phone, plus Exa Connect provider fees). Also note that outputSchema adds a synthesis step per request, which affects user experience even though it does not change the base request price.

Conclusion

The cost of building financial research infrastructure on a web-data API is mostly a function of query volume, and the published numbers make it modelable: roughly $70 per 10,000 standard searches, $12 per 1,000 deep searches (with synthesis when you add outputSchema), and $0.012 to $1.00 per fixed-effort agent run, with the default auto mode and ultra usage-metered under default per-run caps of $5 and $20. Exa's free tier includes $10 in monthly credits plus a one-time $10 onboarding bonus with no payment method required, so you can measure your actual usage pattern before committing. For teams whose volume is spiky or whose tooling needs are custom, that usage-based profile is often the more economical path; for constant, heavy interactive use, run the math against a seat-based subscription before deciding. You can start testing on Exa's pricing page.