Linkup API Pricing and Cost per 1,000 Calls

Linkup sells a web search API for AI agents, and its price list is short enough to read in a minute. That is the appeal. There is no plan ladder and no credit multiplier table, just a rate per call that depends on which endpoint you hit and which options you switch on.

The options are where budgets go wrong. The same search endpoint costs $5 or $50 per 1,000 calls depending on one parameter, and the same fetch endpoint costs $1 or $10 depending on two. This guide lists every published rate, read on 14 October 2026 from Linkup's pricing page and its documentation, then prices six real workloads so you can see which parameter your bill will actually turn on.

How much does the Linkup API cost?

Linkup bills each successful call against a prepaid balance in US dollars. Search costs $0.005 to $0.055 per call, Fetch $0.001 to $0.011 per page, and Research $0.25 to $2.50 per run. There is no monthly fee and no minimum. Prices moved from euros to dollars at parity in May 2026, so older reviews quoting euro figures show the same numbers under a different symbol.

Two smaller items sit beside the table. An Extract endpoint is in closed beta, priced by crawl complexity, with most completed tasks costing $2 to $10 and a $10 minimum balance to submit one. And agents can pay per request in USDC without an account through the x402 protocol, at the same rates with a one cent minimum per call.

Is the Linkup API free?

Partly. An account created with a professional email address receives $20 of credit at sign-up, and eligible accounts are topped back up to $20 every month. At $0.005 a call that is 4,000 standard searches a month at no cost, which is the figure on Linkup's pricing page. The top-up restores the balance to $20 and does not stack, so unused credit does not grow into a reserve.

The grant was $5 until the May 2026 change, so comparisons written before then understate it by four times. Startups can also apply for $5,000 in credits. The free balance is a fair way to prototype an agent, but it covers only 400 deep searches or 8 XL research runs, so it tells you little about what production will cost.

What does a deep search cost on Linkup?

A deep search costs $0.05 per call for raw results and $0.055 for a sourced answer or structured output, exactly ten times the standard rate. Deep runs several search and read passes before answering, so the price reflects more work, not a different product. The depth is one request parameter, which means a single config change turns a $250 monthly bill into $2,500.

Most agent traffic does not need it. A sensible pattern is to send every query at standard depth and escalate to deep only when the first answer comes back thin. If one query in ten escalates, the blended rate is about $9.50 per 1,000 instead of $50.

Linkup Fetch pricing with and without JavaScript rendering

Fetch reads one URL you already have and returns it as markdown. The base price is $1 per 1,000 pages, among the lowest published rates for a page read. Turning on JavaScript rendering multiplies it by five, to $5 per 1,000. Pro mode costs $5 without rendering and $10 with it, and a schema for structured output adds a flat $1 per 1,000 on top of any of them.

So the honest range for a typed, rendered page is $6 to $11 per 1,000. That matters because most modern product, pricing and listing pages build their content in the browser, and a fetch without rendering returns an empty shell on those. Test your real targets with rendering off before you budget at the $1 rate.

Six monthly workloads priced on Linkup

Each row uses the published per-call rates above. The $20 monthly credit is shown where it changes the answer and ignored where it is noise.

The pattern is clear once it is laid out. Linkup is inexpensive when the unit of work is a question, and it gets expensive when the unit of work is a page that needs a browser. The last row is the one to watch if your agent reads far more than it searches.

Does Linkup give volume discounts?

Not on the rates, but on top-ups. Adding $1,000 at once earns 10 percent extra credit, $5,000 earns 15 percent and $10,000 earns 20 percent, according to the May 2026 changelog. In effective terms that is about 9 percent, 13 percent and 17 percent off. A $10,000 top-up buys $12,000 of calls, which brings a standard search to roughly $4.17 per 1,000. Custom pricing, private indexes, zero data retention and an SLA are sold on an enterprise contract.

What happens when the Linkup balance runs out?

The API returns HTTP 429 until you add credit. That is the same status code most clients treat as a rate limit and retry, so an agent that runs dry in production can spin on retries instead of failing loudly. Check the response body before you back off, and set a balance alert well above zero.

Two rules work in your favor. A call that errors is not charged, and that includes a query that returns no relevant results. And the published rate limit is 10 queries per second per organization, shared between Search and Fetch, which is about 26 million calls a month if you could hold it flat. Throughput is unlikely to be the constraint before budget is.

Linkup vs Tavily vs Exa pricing

All three are search APIs for agents, and at standard depth they land within a few dollars of each other. The differences are in what a search includes and how you pay.

Linkup has the largest free balance of the three and the simplest list. Exa is far cheaper for deep search and includes the text of ten results in the search price. Tavily suits teams that prefer a plan to a balance. The detail behind the other two columns is in our Exa pricing guide, the Tavily pricing breakdown and the head to head Exa vs Tavily pricing comparison.

When a per-call search API is the wrong meter

A search API earns its price when the agent does not know where the answer lives. Plenty of production workloads are the opposite. The sites are known in advance, the job is to read every page on them on a schedule, and the page count is in the hundreds of thousands. Paying a per-call rate with a rendering multiplier for that is how the last row of the workload table reaches four figures.

The same goes for a question you ask on repeat. If a Research run exists to answer "what did our rivals change this week", you are paying $0.50 to $2.50 a time for an open-ended investigation of a closed question, and a dedicated competitor monitoring tool that watches the same pages continuously is usually the cheaper purchase for a marketing team.

For engineering teams that need the pages themselves, a flat page allowance is the alternative. Our own plans count one page per page returned with JavaScript rendering included: $99 a month on Startup for about 250,000 pages, which is $0.40 per 1,000, against $1,250 for the same rendered pages through Linkup Fetch. Schema extraction is included from Startup, up to 900 model-mapped pages a month there and 3,500 on Scale, and pages that already publish the fields as schema.org data are mapped without the model. We run no web search, defeat no anti-bot system and have no free plan, so if your agent starts from a question, buy a search API. If it starts from a list of sites, the web crawler API is the cheaper meter, and Firecrawl vs Exa pricing shows the same split between two better known vendors.

How to choose

• Prototype or low-volume agent: Linkup, since $20 a month of credit covers 4,000 standard searches.
• Cited answers at standard depth: Linkup at $6 per 1,000 is competitive with anything published.
• Heavy use of deep search: price Exa first, at $12 to $15 per 1,000 against $50 to $55.
• Long research reports: Linkup Research is simple to budget at a fixed price per run, so count the runs.
• Reading known sites in bulk with rendering: a flat per-page crawl plan, not a per-call fetch.
• A fixed monthly invoice: a plan-based vendor, because a prepaid balance moves with usage.

Whatever you shortlist, run a day of real agent traffic through it and read the cost per call from the responses. The parameter that sets your bill is rarely the one on the pricing page headline. For the wider field on one basis, see web scraping API pricing compared.