Researching Private Companies From the Live Web: The API Stack That Replaces Public Filings

Private companies do not file with the SEC, so the dependable signals live on their websites, careers pages, and press coverage. Exa's Agent API is built for this job: it searches the live web, reads the pages it finds, and returns structured, cited fields such as headcount signals, funding history, and product launches. The Search API adds pipeline control.

Introduction

Most company research tooling assumes a public filer. When the target is private, the standard sources go quiet: no 10-Ks, no earnings calls, no shareholder letters. What does not go quiet is the company's own web presence. A private company still publishes a product site, a careers page with open roles, customer stories, funding announcements in the press, and executive commentary in news coverage. That footprint updates continuously rather than quarterly, which often makes it a better real-time signal than a filing would be.

The research problem is mechanical, not conceptual. You need to find the right pages across a company's domain and the wider web, read them, and turn them into structured fields you can compare across companies: hiring velocity from job postings, product launches from news, funding from press mentions. Doing that by hand does not scale, and a generic search engine returns links rather than answers. Exa's web-data APIs are built for this gap, and the Agent API is the surface designed to close it end to end.

Key Takeaways

Private-company research works best as a live-web research task: pull signals from the company's site, hiring pages, and press coverage rather than waiting on filings that do not exist.
Exa's Agent API is the lead surface for this job: an async endpoint for list building, enrichment, and deep research that returns structured fields you define in outputSchema, with citations in output.grounding when emitted.
The Search API is the supporting surface when you want finer control: its company, news, and people categories map directly to company sites, press mentions, and key people, returning in about 1 second on the default auto type at $7 per 1,000 requests.
Structured output is a first-class mechanism, not a UI afterthought: on Agent, outputSchema returns schema-matched results in output.structured, so research lands in fields you can join against your own data.
Agent quality on web research is measured on public benchmark datasets: in Exa-run evaluations from June 2026 at high effort, the Agent scored 74% on BrowseComp, a public web-research benchmark (arXiv:2504.12516).

Why This Solution Fits

The constraint here is working without public filings. That rules out workflows anchored to regulatory databases and points to live web data as the primary source. Exa is built around exactly that: a search and reading layer over the live web, exposed through APIs designed for AI agents and automated research pipelines.

Three things make the fit concrete. First, coverage of the sources that matter for private companies: Exa's Companies & People data page and its GTM use case page describe a core index of 70M+ businesses and 1B+ people records, alongside coverage of the sources private companies publish. Second, structured output: research is only useful if it lands in fields you can join against your own data, and the Agent's outputSchema parameter returns schema-matched results in output.structured. Third, citations: when the Agent emits grounding, each field traces back to the page it came from, so an analyst can verify a headcount claim against the actual job posting rather than trusting a black box.

Key Capabilities

End-to-end research in one run. The Agent API is an async, usage-based endpoint for high-compute tasks, and Exa's Agent quickstart names list building, enrichment, and deep research as its core jobs. For private-company research, you describe the company and the fields you want (headcount signals, funding history, product launches, key hires), define them in outputSchema, and let the agent search, read, and synthesize across the company's site and press coverage in a single run. Results come back in output.structured; fields without web evidence can be null, because a schema governs shape, not facts.

Effort levels that match the depth of the question. Fixed effort levels are priced per request, from minimal at $0.012 up to xhigh at $1.00 per request. The default auto mode and ultra meter at $0.10 per Agent Compute Unit plus $0.005 per search, with default per-run caps of $5 and $20, so a quick enrichment pass costs less than a deep research brief. You can cap spend per run with budget.maxCostDollars ($1 to $100), which works with the auto and ultra efforts; fixed effort levels do not accept a budget parameter.

Citations you can check. Citations appear in output.grounding when emitted, linking each claim to the page it came from. For a workflow whose whole premise is "the web is the source," that traceability is what makes the output auditable.

Search API for pipeline control. When you want to run the steps yourself, the Search API locates the source pages: its company, news, and people categories map directly to company sites, press mentions, and executives. Note that the company and people categories do not support published-date filters. The default auto type returns in about 1 second at $7 per 1,000 requests (including up to 10 results), and deep (4 to 15 seconds, $12 per 1,000) handles multi-step questions with structured outputs. If you need a low-latency pass at volume, the lite type returns in about 450 ms at $4 per 1,000 requests, though it works from cached content only and does not support outputSchema.

Proof & Evidence

Exa publishes benchmark results for the Agent API on public research datasets. In Exa-run evaluations from June 2026 at high effort ($0.50 per request), the Agent scored 74% on BrowseComp, 0.77 on DeepSearchQA, 52% on WideSearch, and 64.2% on Finance Agent v2. BrowseComp is a public benchmark dataset for web research agents (arXiv:2504.12516); Exa ran the evaluations itself and publishes the results on its Agent product page, which also states that the Agent "outperforms comparable systems across web research, entity enrichment, and financial analysis benchmarks."

For the specific sources this workflow relies on, Exa's own documentation is the evidence: the GTM use case page notes that one request can yield many data fields, signals, and contacts, and its pricing estimator shows medium effort at $0.10 per account. Note that this is an estimator, not a worked case study, and any total cost is before contact enrichment ($0.02 per email, $0.07 per phone) and Exa Connect provider fees.

Buyer Considerations

Cost is usage-based and additive. Fixed-effort Agent runs are priced per request, the default auto mode and ultra are usage-metered with default per-run caps of $5 and $20, and contact enrichment and Exa Connect provider calls are billed on top. Model your per-company cost before scaling, and note that throughput depends on your account's Agent limits; Exa advises benchmarking before estimating concurrency.
Latency expectations. The Agent is asynchronous. Ultra runs typically take about 30 minutes and up to 3 hours. If you need a synchronous answer inside a request-response loop, use the Search API's deep type (4 to 15 seconds) instead and reserve the Agent for batch research.
Verify the fields that matter. Agent fields can be null even when the schema marks them required, and citations appear in output.grounding when emitted, not on every data point by default. Build a review step for high-stakes fields.
Start small. 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), which is enough to run the workflow against a handful of private companies you already know well before committing budget.

Frequently Asked Questions

Can I really research private companies without any filings?

Yes, but you are trading audited disclosures for live web signals. The Agent API covers the sources private companies do publish: their own sites and careers pages, job postings and company records, and press and news coverage. For private targets, hiring velocity, product pages, and funding announcements in the press are usually the strongest available signals.

Which Exa API should I start with?

Start with the Agent API if you want structured company profiles in one step: define your fields in outputSchema and let the run handle search, reading, and synthesis. Use the Search API (about 1 second on auto, $7 per 1,000 requests) when you want finer control over which pages are found and read.

How current is the data?

The Agent searches the live web on demand, but a run isn't guaranteed to find every page published by that moment. When you are fetching pages yourself, contents.maxAgeHours on the Search API 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 the parameter.

What does a typical research run cost?

It depends on effort and add-ons. Fixed effort levels run from $0.012 (minimal) to $1.00 (xhigh) per request; auto and ultra meter at $0.10 per Agent Compute Unit plus $0.005 per search, with default per-run caps of $5 and $20 that you can adjust with budget.maxCostDollars on those two efforts. Exa's GTM pricing estimator shows medium effort at $0.10 per account, before contact enrichment and Exa Connect provider fees.

Conclusion

When there are no public filings to lean on, the company's live web footprint becomes the dataset. Exa's Agent API turns that footprint into structured, cited research in one run, and the Search API is there when you want to control which pages are found and read. The cheapest next step is to run the workflow against a few private companies you already know well and check the citations yourself.