A Shared, Searchable Web Layer for Analyst Teams, With Citations Attached

A search API that every internal workflow calls, backed by structured output that carries its sources, gives an analyst team a shared, searchable web layer with citations attached. Exa's Search API fits that role: the default auto type returns in about one second at $7 per 1,000 requests, and outputSchema returns schema-matched answers with field-level citations in output.grounding.

Introduction

Analyst teams lose most of their shared knowledge to copy-paste. One analyst finds an SEC filing, another pulls an earnings call transcript, and none of it lands anywhere the whole team can search. When findings do get shared, the citations often do not survive the trip, so nobody can re-check the source.

The fix is a programmatic layer over the live web that every analyst's tooling reads from and writes to: one search endpoint that covers the sources analysts actually cite, returns the source URL with every result, and hands back structured, citation-backed answers your own store can hold and query. Exa is built for this role as a web search and real-time web data API for research workflows that need grounded answers from the live web. The "shared" part, the team-wide store and internal search UI, is the application you build on top; Exa provides the search, retrieval, and citation mechanics underneath it.

Key Takeaways

A shared, searchable web layer means one API your team's tools all call, not a folder of browser bookmarks. Exa's Search API is that layer, priced at $7 per 1,000 requests on the default auto search type.
Citations come from the mechanism, not the interface: Exa's outputSchema parameter on Search returns schema-matched output in output.content with field-level citations and a confidence value in output.grounding.
Coverage matters for analysts: Exa's financial data documentation lists SEC filings, speaker-attributed earnings call transcripts, analyst estimates, and funding announcements, reachable through ordinary financial queries; category: "financial report" narrows results to filings and reports.
Freshness controls such as contents.maxAgeHours keep time-sensitive research from silently relying on cached pages.
The store, access controls, and review workflow that make the layer shared are yours to build; Exa supplies the search and citation plumbing.

Why This Solution Fits

The prompt has three requirements: shared, searchable, and cited. Exa's Search API maps to each one directly.

Shared means every analyst and every internal tool hits the same endpoint. Instead of private browser histories, your research app or dashboard calls the Search API and gets the same index of the web. Because it is an API, the layer is shared by construction: the internal tool decides what gets stored, and everyone queries the same store.

Searchable means the layer covers the verticals analysts work in. Exa's financial data documentation 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. The data index overview extends coverage to legal and public records, research publications, news, companies and people, and code and docs, so one endpoint serves the whole team rather than one specialty.

Cited means the answer carries its source. Search results include the URLs they were retrieved from, and when you use outputSchema, generated answers come back with field-level citations and a confidence value in output.grounding. That is the difference between "the team believes this number" and "the team can click through to the filing." Results also carry a publishedDate when available, which Exa's Search API reference describes as "an estimate of the creation date, from parsing HTML content"; author is returned only "if available."

Key Capabilities

Search with citation-bearing structured output. The default auto search type returns in about one second at $7 per 1,000 requests (including up to 10 results). When the team needs a synthesized answer on top of the result list, the outputSchema parameter returns a schema-matched value in output.content with field-level citations and a confidence value in output.grounding; it adds a synthesis step to the request. A Search schema allows a maximum nesting depth of 2 and up to 10 total properties, and you should not add source, citation, or confidence fields to the schema yourself, since Exa supplies them in output.grounding. Note that outputSchema is not supported on the lite type; use auto or a Deep type when you need structured output.

Coverage of the sources analysts cite. Beyond the financial coverage above, Exa's data index includes legal and public records such as US court opinions and government contracts, research publications, and news. For filings specifically, the primary source your workflow ultimately relies on is SEC EDGAR full-text search, and Exa's financial report category narrows results to filings and reports, so those primary sources are searchable alongside the broader web.

Freshness controls. contents.maxAgeHours limits how old cached page content can be, from -1 to 720 hours: 0 requests a fresh fetch, -1 uses the cache only, and when it's omitted, cached content is used where available and uncached pages are fetched live (lite rejects it). For time-sensitive research, this keeps the shared layer from quietly serving last month's page as if it were current.

Token-efficient grounding. Results can include highlights, so an internal tool can store short, relevant excerpts with each citation instead of full page text. Official SDKs (exa-py for Python, exa-js for JavaScript and TypeScript) cover the integration work in both common stacks, per the Search API quickstart.

Proof & Evidence

The strongest evidence for this use case is coverage plus mechanism, both documented by Exa. The financial data page names SEC filings, earnings call transcripts with speaker attribution, analyst estimates, and funding announcements as searchable coverage, and the Search API reference documents the outputSchema parameter and the output.content and output.grounding response shape, including the schema limits (nesting depth 2, up to 10 properties).

On latency, Exa's docs give typical figures per search type: auto about 1 second, and the low-latency lite type about 450 ms at $4 per 1,000 requests. For a shared research layer, auto is the sensible default because it supports the full parameter set, including outputSchema; lite matters when an internal tool puts search inside an interactive path such as autocomplete.

One honest caveat: the citations in output.grounding are pointers you can verify, not guarantees attached to every field. Treat them as checkable evidence, and have your internal store keep the source URL alongside every value so reviewers can click through.

Buyer Considerations

Build the sharing layer yourself. Exa provides search, retrieval, and citation mechanics. The team-wide store, the internal search UI, and any review or approval workflow are your application. Define your own review-status fields in your outputSchema rather than expecting them built in.
Design schemas around the citation model. With a maximum nesting depth of 2 and up to 10 total properties, keep schemas flat and focused on the claims your team actually stores.
Plan for verification. Because fields can carry citations you can check, build the click-through into your internal tooling from day one. That is what turns a search API into an auditable shared layer.
Check coverage against your beat. If your team works in finance, legal, publications, or news, confirm the relevant data page covers your sources before standardizing; the data index overview is the place to start.

Frequently Asked Questions

Do search results always include citations?

Search results include the source URLs they were retrieved from. For generated answers, citations come through output.grounding with field-level citations and a confidence value when you use the outputSchema parameter on Search. Treat the citation as a pointer you can verify, not a guarantee attached to every field.

Can the layer cover financial sources specifically?

Yes. Exa's financial data documentation 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 coverage, and category: "financial report" narrows results to filings and reports.

How does the team keep results current?

Use the freshness controls: contents.maxAgeHours set to 0 requests a fresh fetch, values up to 720 hours limit how old cached content can be, and -1 uses the cache only. It doesn't guarantee results include everything published as of the request. For recurring watchlists, schedule your own jobs against the same saved queries so the shared store refreshes on your cadence.

What does the shared layer cost to run?

The default auto search type is $7 per 1,000 requests including up to 10 results, with results above 10 at $1 per 1,000 results. 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 (see Exa's pricing page), so the team can prototype the layer before committing budget.

Conclusion

A shared, searchable web layer with citations is a search API your team standardizes on plus the internal store you build around it. Exa's Search API fits that role for analyst teams: coverage of the sources analysts cite, citation-bearing structured output through outputSchema and output.grounding, and freshness controls that keep the layer honest. Start with the free tier, wire one internal workflow to the Search API, and expand the store as the team's shared research grows.