List building · Updated
Exa review (2026): how well its AI agent builds lead lists
Exa is a capable general-purpose list builder: describe a list in plain language and its agent searches a large web index for a few hundred accurate rows. It is usually the weaker choice when a filter-based search fits the list, or when the full list is public on one page and can be scraped. Its niche is narrow: lists no filter can describe.
pipe0 buys from Exa and offers its Agent as a list-building search. The scores come from our own benchmark, not from that relationship. How we review
What we reviewed
Exa is a search engine built for AI: web search, page contents, answers, research agents and Websets. We reviewed it for one job: building prospect lists for sales and marketing teams from a plain language prompt.
Tested for
- Local business lists with the Agent API (ultra and medium effort)
- Local businesses through search's places results
- Online brands and trade-show exhibitors, scored against ground truth
Not scored here
- Web search and contents for RAG and AI apps
- Websets, Exa's list product, which has its own paid plan
Our score for Exa, the average of four criteria
The best prompt-based list builder we tested: a few hundred accurate rows in minutes, ahead of Parallel every time. When a database filter can describe the list, the filter wins.
Best for
- A few hundred accurate rows from a prose brief no filter can express
- Lists published on the web, like exhibitors or members, when a sample is enough
- Agents that need list building behind one tool call
Skip it if
- You need thousands of local businesses per city
- You need every entry of a directory, not a sample
- A database filter can describe your list (industry, size, tech stack)
- Your workflow can't wait two to six minutes for a result
A prompt and an output schema in, a structured list out. Runs are asynchronous and take two to six minutes, and search's 10-place cap isn't in the docs.
Rows were mostly precise, 96% on exhibitors, but lists were short (146 to 341 local businesses) and phone coverage swung from 97% to 8%.
A medium run is $0.10 flat. Ultra is metered and cost us about $1.10 per run, so short lists landed at $3 to $8 per 1,000 rows.
A hosted MCP server with an agent tool, Python and JavaScript SDKs, spend and time caps per run, and up to 50 runs at once.
Exa at a glance
- Agent API
- Asynchronous runs from a prompt and an output schema; effort from minimal to ultra Source
- Run limits
- Budget $1 to $100 per run; ultra time limit 5 minutes to 3 hours Source
- Places
- Local business profiles with address, hours and reviews, served inside search; 10 per call in our tests Source
- Rate limits
- Search 10 requests per second; Agent 50 active runs Source
- MCP server
- Hosted at mcp.exa.ai/mcp, with search, fetch and agent tools Source
- SDKs
- Python and JavaScript Source
What is Exa?
Exa is a search engine built for AI applications. Its API searches the web, returns page contents, answers questions, and runs research agents. For sales and marketing teams the relevant part is the Agent API: describe a list in plain language, define the columns, and it searches, checks and returns structured rows.
Can Exa build lead lists?
Yes, lists of a few hundred rows. An Agent run returned 146 to 341 local businesses per city in our tests, and the rows were accurate. It beat Parallel's research tasks on every list. Purpose-built data beat it on volume: DataForSEO's business listings returned two to six times as many local businesses, in seconds.
How many businesses each tool finds
Unique rows per list, one run each. All four panels share one scale, so a short bar is a short list.
Barbershops in New York City
DataForSEO
Exa ultra
Exa medium
Parallel Task pro
Parallel Entity
Crustdata
Amplemarket
HVAC contractors in Dallas–Fort Worth
DataForSEO
Exa ultra
Exa medium
Parallel Task pro
Parallel Entity
Crustdata
Corporate event venues in Chicago
DataForSEO
Exa ultra
Exa medium
Parallel Task pro
Parallel Entity
Crustdata
Amplemarket
Property managers in Miami
DataForSEO
Exa ultra
Exa medium
Parallel Task pro
Parallel Entity
Crustdata
Amplemarket
Company databases count every row they returned; most were neighbouring businesses such as suppliers, software and planners.
A run can return far more when the list already exists in one public source. Asked for New York City barbershops, Exa found the state licence registry and returned 1,932 rows, 1,715 of them licensed shops. Registry rows carry legal entity names ("CUTTING POINT CORP") and rarely a phone number: 8% had one, against 80% in DataForSEO's list.
New York City barbershops against the licence list
New York State licenses 1,952 barbershops in the city, marked by the line.
Checked against the licence list
matched by address
DataForSEO
80% with a phone, all inside NYC ZIP codes
Exa ultra
copied the licence registry; 8% with a phone
Not checked
row count only
Exa medium
all inside NYC, 68% with a phone
Parallel Task pro
Parallel Entity
companies, no addresses
Parallel Task ultra
55% inside NYC
Amplemarket
about 18 real
Crustdata
about 8 real
A list can pass the licence count: unlicensed shops, salons that also cut hair, and permanently closed listings, which DataForSEO flags.
How long does an Exa Agent run take, and what does it cost?
An ultra run takes three to six minutes and costs about $1 to $1.50; a medium run takes about two minutes for a flat $0.10. Both are slower than a database query and faster than Parallel's research tasks, which took eight to fourteen minutes. Ultra's time limit can't be set below five minutes, so it doesn't suit a step that must answer quickly.
How long a list takes
Time per run, from each vendor's own timestamps. Each bar spans the fastest and the slowest list.
Filter and company search
DataForSEO
Filter search over map listings
Parallel Entity
Company search
Prompt-based agents
Exa medium
Prompt-based research agent
Exa ultra
Prompt-based research agent
Parallel FindAll
Verified list search
Parallel Task pro
Prompt-based research task
Parallel Task ultra
Prompt-based research task
The line marks 10 minutes. Past it, a run doesn't fit an interactive workflow or the step limit most automation tools set.
Because you pay per run, the cost per row depends on how many rows the run finds: $0.46 to $1.23 per 1,000 rows for medium and $0.76 to $7.53 for ultra, against $0.37 for DataForSEO.
What 1,000 rows cost
Vendor price per 1,000 rows, before any reseller margin. Log scale: each tick is ten times the last.
Filter and company search
DataForSEO
Filter search over map listings
Parallel Entity
Company search
Prompt-based agents
Exa medium
Prompt-based research agent
Exa ultra
Prompt-based research agent
Parallel FindAll
Verified list search
Parallel Task pro
Prompt-based research task
Parallel Task ultra
Prompt-based research task
Per-run pricing makes a short list expensive per row. Parallel Entity is the cheapest per row, but its rows are company profiles without an address or a phone.
When does a database filter beat Exa?
Whenever a filter can describe the list. Online-only brands look like Exa's territory, since they have no storefront to find on a map. But a company database with a technology filter found three times as many verified Shopify brands in seconds, and already held 25 of the 26 Exa found.
Online brands: a filter beats a prompt
“US sustainable apparel brands with their own Shopify store.” A brand counts as verified when its domain serves a live Shopify store.
Verified brands per search
cost in pipe0 credits
Crustdata, Shopify technology filter
seconds · 13.7 credits
Amplemarket, first page
seconds · 2 credits
Exa Agent, ultra
165 s · about 65 credits
Crustdata already held 25 of the 26 brands Exa verified, and lists 5,726 US apparel companies with Shopify detected.
The same holds for local businesses with a clear category. Use Exa for what a filter can't express, then check whether the database had it first.
Can Exa get every exhibitor or member from a public list?
No. It returns an accurate sample of a few hundred, not the whole list. On a trade show whose 3,194 exhibitors are public on one page, Exa returned 212, and 96% of them were real. A research agent that can read the directory page directly returned the full list, but only when it found that page; a typical run returned fewer rows than Exa.
A public directory: how much of it each tool returns
Exhibitors at Natural Products Expo West 2026. The official list of 3,194 sits on one public page; each track is the full list.
Exhibitors returned
share of the official list
Exa Agent, ultra
about 4 min · about 63 credits · 96% precise
Research agent, typical run
1 to 3 min · a few credits
Research agent, reached the directory page
about 3 min · 8 credits · 1 run in 9
An agent that writes every row itself tops out at a few hundred. Copying the whole list needs a tool that reads the directory page directly, and finding that page is the hard part.
So Exa is the dependable choice when a few hundred correct rows are enough. When you need every entry, point a scraper or a list-reading agent at the directory page. We haven't yet tested lists that no single page holds, such as companies mentioned across many news articles.
Can Exa's search return local businesses?
Yes, 10 at a time. Exa's search returns places with address, phone, rating and opening hours, and they were precise: 591 of 602 NYC barbershops were inside the city. Each call returns at most 10 places, so a city-wide list means splitting the city into ZIP codes yourself. Across 178 NYC ZIP codes that gave 602 unique barbershops in 20 seconds for $1.25, about 15% of what DataForSEO returned.
Exa and pipe0
pipe0 offers Exa's ultra agent as a list-building search, with a $5 spending cap and a 10-minute limit per run, billed from Exa's own reported cost. Its agent sends local businesses to DataForSEO and filterable companies to the company databases, and uses Exa for briefs no filter can express.
Exa, DataForSEO or Parallel: which should you use?
Use DataForSEO for local businesses by category, a company database for anything a filter can describe, and Parallel for researching rows you already have. Use Exa when the list needs a prose brief and a few hundred accurate rows are enough.
Which tool for which list
Every cell cites the measurement behind it. Company databases are Crustdata and Amplemarket, searched with their filters.
| Job | DataForSEO | Exa | Parallel | Company databases |
|---|---|---|---|---|
| Local businesses by categoryBarbershops in New York City | Best pick826 to 3,959 rows per city in seconds | Works146 to 341 rows in 3 to 6 minutes | Weak17 to 200 rows in 8 to 14 minutes | Weak26 barbershops, most of them noise |
| Phone, hours and rating per locationHVAC contractors to call | Best pick80 to 96% of rows with a phone | Mixed97% with a phone in Dallas, 8% in NYC | WeakEntity Search returns no phone | Not built for itNo phone per location |
| Companies by firmographics or tech stackUS apparel brands on Shopify | Not built for itOnly businesses with a map listing | Weak26 brands, 25 already in Crustdata | Not testedEntity Search wasn't run on this | Best pick77 verified Shopify brands in seconds |
| Lists published on the webExhibitors at a trade show | Not built for itNo event or membership data | WorksA few hundred accurate rows per run, never the full list | Not testedFindAll fits the shape; not run | Not built for itNo exhibited-at filter |
| One researched answer per rowThe imprint page of a small company | Not built for itNot a web research tool | Not testedNot benchmarked for this | Best pickTask base solved 5 of 5, in 22 to 63 s | Not built for itNot a web research tool |
One run per scenario. “Not tested” means we didn't measure it, not that it fails.
Exa pros and cons
Pros
Beat Parallel on every list
On four local lists, Exa's ultra agent returned 146 to 1,932 rows in three to six minutes. Parallel's tasks returned 17 to 200 in eight to fourteen.
Accurate rows from a prompt
Rows were real and on target: 96% of the trade-show exhibitors it returned were on the official list, and every NYC row from a medium run was inside the city.
Built for agents
A hosted MCP server, budget and time caps per run, and a structured output schema make it a safe tool to hand an agent.
Cheap fixed runs
Medium effort costs $0.10 a run and returned 81 to 217 local businesses in about two minutes, every NYC row inside the city.
Cons
A few hundred rows per run
Unless it finds a registry to copy, a run stops at a few hundred rows: 146 to 341 local businesses, or 212 of a show's 3,194 exhibitors. DataForSEO returned 826 to 3,959 local businesses.
Slow for interactive work
Agent runs took two to six minutes, and ultra's time limit can't go below five minutes.
Uneven contact data
97% of Dallas HVAC rows had a phone number, but 8% of NYC barbershops did, and 39% of Miami rows fell outside Miami-Dade.
Places search caps at 10
Search returns at most 10 places per call. Wider lists mean splitting a city into ZIP codes yourself.
Exa pricing
Exa has a free plan. Paid plans start at Free pay as you go, with $10 in free credit each month.
| Plan | Price | What you get |
|---|---|---|
| Free credit | $10 a month | Resets on the 1st of each month. No subscription or minimum spend. |
| Search | $7 per 1,000 requests | Fast and auto search with up to 10 results; deep search $12, extra results $1 per 1,000. |
| Agent, fixed effort | $0.012 to $1.00 per run | Minimal, low, medium ($0.10), high and xhigh effort, each at a flat price. |
| Agent, auto and ultra | Metered | $0.10 per compute unit plus $0.005 per search, capped by a $1 to $100 budget per run. |
| Websets | Separate plan | Exa's list product, billed apart from API credit. |
- Ultra runs default to a $20 spending cap and $5 for auto; we used ultra with a $5 cap and a 10-minute limit.
- Contact enrichment inside an Agent run costs $0.02 per email and $0.07 per phone number.
Prices as of Oct 15, 2026, from Exa's pricing page.
Exa alternatives
| Alternative | Pick it when |
|---|---|
| pipe0 list-building search | You want Exa's agent in a sheet or an MCP workflow, billed from Exa's own cost, with enrichment on the rows it returns. |
| DataForSEO | You need local businesses by category: thousands of rows per city in seconds. |
| Crustdata | Filters can describe your companies, including the tech they use. |
| Parallel | You need one researched answer per row rather than a new list. |
Frequently asked questions
Can Exa build lead lists?
Yes, lists of a few hundred rows. Exa's Agent API turns a prompt into a structured list. In our tests an ultra run returned 146 to 341 local businesses per city in three to six minutes for about $1.10.
Can Exa extract a full exhibitor list or directory?
No. On a trade show whose 3,194 exhibitors were public on one page, Exa returned 212 of them, 96% correct. To get every entry, use a tool that reads the directory page itself.
How much does an Exa Agent run cost?
Fixed efforts cost $0.012 to $1.00 per run; medium is $0.10. Auto and ultra are metered at $0.10 per compute unit plus $0.005 per search, within a budget you set between $1 and $100. Our ultra runs cost $1.09 to $1.46.
How many places does Exa search return?
In our tests, 10 places per search call, whatever numResults was set to. Exa recommends the Agent for longer lists.
Exa or Parallel for building lists?
Exa. On four local business lists Exa's ultra agent returned more rows than Parallel's Task API every time, in three to six minutes against eight to fourteen.
Exa or DataForSEO for local businesses?
DataForSEO. It returned two to six times Exa's rows in seconds, at $0.37 per 1,000. Exa found some businesses DataForSEO missed, so for a complete territory you can merge both.
Next-gen enrichment & search.
Build revenue systems that scale. For humans, agents, and apps. Replace tools like Clay, n8n, Hightouch, etc.
Try pipe0