Linkup vs Exa Pricing per 1,000 Searches
Linkup charges $5 per 1,000 standard searches and Exa $4 to $7 with result text included, while deep search is $50 on Linkup against $12 to $15 on Exa. Both price lists and five agent workloads priced on each.
By the ClawEngine team
October 2026 · 7 min read
Hit Extract to turn this page into clean, LLM-ready data.
robots.txt respected · public data only ·
Short answer: Linkup is cheaper for plain standard search and Exa is cheaper for everything deeper. A standard Linkup search costs $5 per 1,000 calls, against $7 for an Exa auto or fast search, though Exa Instant is $4 and Exa includes the text of up to 10 results in that price. Deep search reverses the order: $50 per 1,000 on Linkup against $12 to $15 on Exa. Linkup gives $20 of free credit a month and Exa gives $10. If most of your agent's budget goes on reading pages it already knows, a crawl plan priced per page beats both.
Linkup and Exa get compared because they sell the same thing to the same buyer: a search API an AI agent can call and get back sources it can quote. Both bill per call against a balance, with no monthly plan to pick, and both publish their rates. What they charge for is different. Linkup prices each endpoint by depth and output type, so the same search call costs $0.005 or $0.055 depending on one parameter. Exa prices by search mode and counts content per page, but folds the text of the first 10 results into the search price.
That makes the headline numbers misleading in both directions. This post lists both price lists as each vendor publishes them, read on 19 October 2026 from exa.ai/pricing and Exa's pricing reference, and from linkup.so/pricing and Linkup's pricing documentation, then prices five agent workloads on each.
Linkup vs Exa pricing at a glance
| What you compare | Linkup | Exa |
|---|---|---|
| How you pay | Prepaid balance in USD, charged per call | Pay as you go, no subscription or minimum spend |
| Standard search | $5 per 1,000 with raw results, $6 with a sourced answer or structured output | Instant $4, fast or auto $7 per 1,000, up to 10 results with their text included |
| Deeper search | Deep $50 per 1,000, $55 with an answer | Deep-lite and deep $12, deep-reasoning $15 per 1,000 |
| More than 10 results | Same price per call | $1 per 1,000 extra results |
| A direct answer | Sourced answer on search, $6 per 1,000 | Answer endpoint, $5 per 1,000 |
| Reading a known page | Fetch $1 per 1,000 without rendering, $5 with it, pro mode $5 to $10, a schema adds $1 | Contents $1 per 1,000 pages for each content type, text, highlights or summary |
| Research and agent runs | Research fixed per run, S $0.25, M $0.50, L $1.50, XL $2.50 | Agent fixed per effort, $0.012 to $1.00, or metered with a $5 default cap |
| Free each month | $20 of credit on a professional email, topped back up to $20 | $10 of credit, reset on the first of the month, plus a one-time $10 onboarding bonus |
| Site crawling | No crawl endpoint. Fetch one URL at a time | No crawl endpoint. Contents takes a list of URLs |
Two lines in that table decide most bills. The first is what a search returns. A Linkup search with raw results gives you ranked sources with snippets of content. An Exa search at $7 per 1,000 returns up to 10 results and, if you ask for it, the text of those pages at no extra charge. Exa's pricing reference says each endpoint has a base price per request that includes up to 10 results, and only extra results and Exa-generated summaries are billed on top. So the Exa price is higher on paper and often lower once you count the page text an agent needs anyway.
The second is depth. Linkup's deep mode runs several rounds of searching before it answers, and it costs ten times its standard search. Exa's deep modes cost under twice its auto mode. If your agent defaults to the deepest setting, that one choice is worth more than every other line in the table.
Five agent workloads priced on Linkup and Exa
Each workload below uses list prices only, before the free monthly credit, and assumes the cheapest mode that does the job. Where the two products are not equivalent we say so rather than force a match.
| Monthly workload | Linkup | Exa |
|---|---|---|
| 10,000 standard searches, snippets are enough | $50 | $40 on Instant, $70 on auto |
| 100,000 searches, full text of the top 3 results needed | $500 search plus $300 for 300,000 fetches without rendering, $800 | $700 on auto, text included |
| 20,000 deep searches | $1,000 | $240 on deep, $300 on deep-reasoning |
| 1,000 research reports of medium size | $500 on Research M | $100 on Agent at medium effort |
| Reading 200,000 known pages | $200 without rendering, $1,000 with it | $200 for text, $400 with highlights too |
Read the first row with care. At small volumes the free credit changes the answer: Linkup's $20 covers 4,000 standard searches, Exa's $10 covers 2,500 Instant searches, so a prototype at a few thousand calls a month can run free on either. Linkup wins the cheap end because its monthly credit is twice as large.
The second row is where the headline misleads. Linkup's standard search is $2 cheaper per 1,000 than Exa auto, but if the agent then needs the page text to ground an answer, Linkup charges for each fetch while Exa already returned it. We assumed three fetches per search without JavaScript rendering. With rendering, Linkup's fetch cost rises fivefold and the gap widens.
The fourth row compares two different products. Linkup Research and Exa Agent both run a multi-step job and return a report, but they are tuned differently and their outputs are not interchangeable. The point is the pricing model: both now publish a fixed price per run by size or effort, so a product sold at a fixed price per report can budget either one. Exa also offers a metered mode with a default cap of $5 per run, which is the setting to watch.
Linkup vs Exa on the parts the price list does not show
How results are found. Exa runs its own index and is built around semantic retrieval, which suits open-ended research and "find pages like this one" queries. Linkup positions itself on factual lookups with fresh sources. Neither claim is something a price list can settle, so run 50 of your real queries through both before you commit, and grade the answers rather than the latency.
What failure costs. Linkup does not charge for failed calls, and a search that finds nothing relevant is not billed as a failure, it is a successful call with few results. Exa charges per request regardless of how useful the results were. For agents that fire many speculative queries, count this.
Running out. Linkup returns HTTP 429 when the balance is empty, so a fleet of agents needs auto top-up or a balance alert. Exa bills pay as you go against a card, which removes the hard stop and adds a bill you have to watch. A team that already reconciles several metered AI vendors each month will want the usage exports from both.
When neither search API is the right meter
Search APIs charge to find pages. Many agents spend most of their budget reading pages they already know: a competitor's pricing page, a documentation site, a set of product listings checked every week. There is nothing to search for in that work, and paying a search price or a per-fetch price for it adds up. Reading 200,000 rendered pages costs $1,000 on Linkup Fetch and $200 on Exa Contents if Exa has the pages.
A crawl plan billed per page does that job for less. On ClawEngine the Startup plan is $99 a month for about 250,000 pages with JavaScript rendering included, about $0.40 per 1,000. Our web crawler API walks a site from one start URL, and the web data API for AI agents page shows how teams split the work: a search API to find sources, a crawler to read the sites they keep coming back to. If the agent watches the same pages for changes, the website change monitoring API bills those checks per page too. And when the whole job is an agent that repeats the same lookup every morning, running it on a schedule for AI agents rather than on demand keeps the search bill to the runs you actually need.
To be plain about the limits: ClawEngine has no search product and no semantic index. It cannot replace Linkup or Exa for the "find me sources on this question" half of an agent, and it does not read sites that block automated access. It replaces the reading half on public sites.
How to choose between Linkup and Exa
- A prototype or a small internal agent: Linkup. The $20 monthly credit covers 4,000 standard searches, against 2,500 Instant searches on Exa's $10.
- A production agent that needs only snippets: Exa Instant at $4 per 1,000 if its quality holds on your queries, otherwise Linkup at $5.
- A production agent that grounds answers in page text: Exa, because the text of up to 10 results is in the search price.
- Heavy deep search: Exa, by about four times at list price. Test answer quality on both first, since the two deep modes work differently.
- Research reports sold at a fixed price: either, both publish a fixed price per run. Compare output quality, not the price.
- Bulk reading of known sites: neither. Use a crawl plan billed per page and keep the search API for finding sources.
For the full price list on each side, see our Exa pricing breakdown and the Linkup API pricing guide. If Tavily is also on the shortlist, Linkup vs Tavily pricing and Exa vs Tavily pricing price those pairs the same way, and Exa alternatives covers the rest of the field.
See ClawEngine turn pages into clean data
Point ClawEngine at any public or permitted site and get back clean markdown, JSON, or typed structured fields in one call. Crawl at scale, render JavaScript, and feed your RAG pipelines and AI agents, robots.txt and Terms of Service respected.