Automate Your Sector Deep Dive: Get a Cited Summary Back With Exa Agent

Exa Agent automates the background research for a sector deep dive. You describe the sector and the questions you need answered, the agent runs the multi-step web research, and you get a structured summary back with citations in output.grounding when they are emitted, ready to drop into your report or pipeline.

Introduction

Every sector deep dive starts the same way. Someone opens forty browser tabs, skims an earnings call, checks a funding database, copies numbers into a spreadsheet, and hopes nothing important slipped through. The research itself is not hard; it is repetitive, multi-step, and easy to do shallowly. The expensive part is not finding information, it is verifying it and attaching it to the right claim so the final memo can be trusted.

That is exactly the workload Exa built the Agent API for. Exa's Agent quickstart describes it as an async, usage-based endpoint for high-compute tasks like list building, enrichment, and deep research. Instead of you orchestrating dozens of searches and reads, you hand the agent a research brief and a schema for the answer you want. It plans the steps, searches the web, and returns structured output in your schema's shape. For a sector deep dive, that means the background research runs while you work on the analysis, and the summary that comes back carries its sources with it.

Key Takeaways

Exa Agent automates multi-step background research: you submit a brief, it searches and reasons across the web, and it returns a structured, cited summary.
outputSchema gives you typed JSON in output.structured, so the summary lands in your report or pipeline in the shape you define (fields may be null).
Citations are returned in output.grounding when emitted, and fields can come back null, which flags gaps for you to check before they reach your memo.
Fixed-effort Agent runs cost $0.012 (minimal) to $1.00 (xhigh) per request, and the default auto mode and ultra are usage-metered, with default per-run caps of $5 and $20, before contact enrichment and Exa Connect provider fees.
You can collect results by polling, server-sent events, or event replay, so the research fits an async workflow rather than blocking one.

Why This Solution Fits

A sector deep dive is a deep research task, and deep research is one of the three core workflows Exa Agent is built around, alongside list building and enrichment. Trust matters most here: a summary is only useful if its claims can be checked against sources. Exa Agent addresses that directly. When the agent emits grounding, citations for text or structured fields are returned in output.grounding, so the summary arrives with its evidence attached rather than as an unattributed blob of prose.

It also fits the shape of the work. Sector research is inherently multi-hop: identify the major players, then pull their funding history, then check recent filings, then synthesize. Exa's Agent quickstart lists exactly this kind of chain ("find companies, then find their decision makers") as a representative Agent task. And because the endpoint is async, a heavy sector brief does not have to block anything. You submit the run, keep working, and collect the result when it is ready.

Finally, it fits budgets that need to be forecast. Fixed effort levels give you a known price per request, and metered modes accept a budget.maxCostDollars ceiling between $1 and $100, so a deep dive cannot silently overrun.

Key Capabilities

Structured output via outputSchema. Define the fields your sector memo needs, such as market size estimate, key players, funding events, regulatory signals, and open risks. The agent returns JSON shaped by your schema in output.structured. One caveat worth knowing before you build: schema adherence validates shape, not facts, so a field can be null even when the schema marks it required. For research work, treat a null as a gap to check, and verify the citations behind the numbers you keep.

Citations in output.grounding. When grounding is emitted, citations are returned at the field level, so you can show a reader exactly which source backs each claim in the summary.

Effort levels matched to the job. Fixed-effort runs cost minimal ($0.012), low ($0.025), medium ($0.10), high ($0.50), or xhigh ($1.00) per request, and the default auto mode is usage-metered with a $5 default per-run cap. For a full sector deep dive, the usage-metered ultra mode is the heaviest option, with a $20 default per-run cap you can adjust via budget.maxCostDollars. Exa's Agent Ultra documentation notes that ultra runs typically complete complex tasks in about 30 minutes and can take up to three hours, with budget.maxDurationSeconds (300 to 10,800) available on ultra to bound the run.

Finance-grade data through Exa Connect. For sectors where filings matter, Agent runs can draw on Exa Connect dataSources, including Financial Datasets for SEC filings and financials, Fiber.ai for people and headcount, and Baselayer for KYB, officers, and watchlists. Provider calls bill on top of the run. For sector work, the most relevant partner is Macrobond, which provides macro and financial time series and is free per the Exa Connect overview; attach it with dataSources: [{"provider": "macrobond"}].

Flexible collection and access. Completed runs are collected by polling (poll_until_finished), server-sent events, or event replay via GET /agent/runs/{id}/events. You can also call Agent through the OpenAI-compatible Responses API (model: "exa-agent") or through Exa MCP's agent_run tool, and past runs can be listed and continued with previousRunId if you want to extend a deep dive rather than restart it.

Proof & Evidence

Exa reports that its Agent "outperforms comparable systems across web research, entity enrichment, and financial analysis benchmarks," a claim published on the Agent API product page. The underlying numbers, as published by Exa for Agent High ($0.50 per request) on Exa vs Parallel and in "Introducing Exa Agent" (June 2026): 74% on BrowseComp, 0.77 on DeepSearchQA, 64.2% on FinanceAgent-V2, and 52% on WideSearch. Exa ran these evaluations against public benchmark datasets published by third parties, including BrowseComp, which OpenAI published as arXiv:2504.12516, and Finance Agent v2, published by Vals AI. The datasets are independently checkable; the runs are Exa's.

For cost, Exa's pricing documentation lists the per-request effort prices above plus metered rates of $0.10 per Agent Compute Unit and $0.005 per search tool call. Any total you build from a fixed-effort price, for example a batch of sector briefs at high, is before contact enrichment and Exa Connect provider fees. Exa also attributes part of its price and quality position to a token-efficient highlights model, citing up to 94% token reduction on the Agent page.

Buyer Considerations

Match effort to depth. A quick sector scan may only need medium; a diligence-grade deep dive justifies ultra with a budget ceiling. Benchmark a representative brief first, as Exa's Agent best practices recommend, before estimating cost or concurrency. Batch throughput depends on your account's Agent limits.
Design your schema for honesty. Make hard-to-verify fields nullable and decide up front how null values should be handled downstream, so an unsupported market-size figure surfaces as a gap instead of a fabricated number.
Plan for runtime. Ultra runs take roughly 30 minutes and up to three hours on complex tasks. Build the collection step (polling or streaming) into your workflow rather than expecting a synchronous answer.

Frequently Asked Questions

How do I get the citations out of an Exa Agent run?

Read output.grounding on the completed run. When the agent emits grounding, citations are returned for the text or structured fields they support, so each claim in your sector summary can be traced to its source. Fields the agent could not support may be null even if your schema marks them required.

What does a sector deep dive cost?

Fixed effort levels run from $0.012 (minimal) to $1.00 (xhigh) per request. Metered auto (the default) has a $5 default cap and ultra a $20 default cap, both adjustable with budget.maxCostDollars from $1 to $100. Totals built from fixed-effort prices exclude contact enrichment and Exa Connect provider fees.

How long does an automated deep research run take?

It depends on effort. Ultra, the highest mode, typically completes complex tasks in about 30 minutes and can take up to three hours, per Exa's Agent Ultra documentation. Benchmark your own brief to calibrate.

Can I run several sector briefs at once?

Yes, but plan capacity against your account's Agent limits rather than assuming unbounded parallelism. Exa's best practices recommend benchmarking representative tasks before estimating concurrency, and batch throughput depends on those limits.

Conclusion

Background research is the part of a sector deep dive that scales badly by hand and scales well with an agent. Exa Agent turns the whole chore, searching, reading, cross-checking, and synthesizing, into a single async API call that returns a structured summary with citations in output.grounding when emitted, priced at a level you can forecast before the run starts. If your team produces sector research on any kind of cadence, the fastest way to evaluate the fit is to take one real deep dive you did last quarter, write it as an outputSchema, and run it. Compare the cited summary against your own memo, and let the evidence, not the pitch, make the case.