What a web search API should cost in 2026

Kaer's Search API is priced by lane: $0.50, $1.00 or $3.00 per 1,000 queries. Prepaid, no monthly minimum, no charge for asking for more results. This post lines that up against every provider you are likely to be comparing us with, shows the mechanics that make headline prices misleading, and says plainly where we are not the cheapest.

Search is now the largest recurring line in most agent workloads. A model call happens once per turn; a research agent can search fifteen times before it answers. So the per-query price stops being a rounding error around your first serious customer, and at a million queries a month it decides whether search costs you hundreds or tens of thousands.

Prices in this market are quoted in ways that are hard to line up: per request, per credit where a request can cost two credits, or folded into a token bill. So, one definition before the table: everything below is what you pay for one thousand ordinary searches, with anything extra called out in the notes rather than buried.

The price list, side by side

Kaer and the first nine rows are from our published rate card, each checked against the vendor's own pricing page on 21 September 2026. The last four were read on 22 September 2026. US dollars per 1,000 searches, excluding tokens.

Two things stand out. The model providers' built-in search tools are the most expensive way to buy search, because you pay a convenience premium for not holding a second key. And the cheap end is no longer only SERP proxies: Parallel's Turbo sits at $1.00 and our web lane at $0.50, while most APIs built for agents still sit between $5 and $8.

The headline price is rarely the price you pay

Three mechanics account for most of the gap between a quoted price and an invoice.

Result-count surcharges. Exa's $7.00 covers ten results; every result past the tenth adds $1.00 per 1,000. Ask for twenty, a normal ask when a reranker sits downstream, and you pay $17.00 per 1,000. Serper does the same in miniature: one credit for up to ten results, two for more. A Kaer query returns up to twenty-five results for the price of one.

Credit multipliers. "Per credit" is only comparable to "per query" when a query is one credit. On Tavily an advanced search is two, so the $8.00 headline is $16.00 for the depth most retrieval pipelines actually want.

Minimums and packs. Tavily's lowest effective rate arrives with a $500-a-month plan, and Serper's $0.30 needs its largest prepaid pack. If your volume is genuinely there, those are honest prices. If it is not, you are pre-buying searches you will not run.

One agent hop, priced end to end

Per-thousand rates hide all of that, so here is one concrete unit of work: a single query that asks for twenty results, roughly one hop of a research agent. The last column is the same hop a million times a month.

Same hop, a spread of $500 to $17,000 a month. That spread, not the headline number, is what belongs in the budget.

Where we are not the cheapest

A comparison that ends with its author winning every row is an advert, so: if ten results are enough and you are buying at serious volume, Serper's largest pack at $0.30 per 1,000 undercuts our web lane. And Parallel's Turbo matches our fast lane at $1.00. We do not chase those prices down, because a SERP proxy is a different thing to build: it fetches another engine's results page on demand and parses it, with no index of its own behind it.

That difference cuts both ways, and it is worth being precise about which way it cuts for you:

• You inherit someone else's limits. A proxy's availability depends on how the upstream engine feels about being proxied that week. Ours depends on our own capacity.
• Latency has a floor. A proxy cannot be faster than the engine it fronts plus a round trip. The web lane answers from our index in ~80 ms.
• What comes back differs. A results page gives you a title, a URL and a snippet. The deep lane adds what an agent needs to decide which pages to open: freshness, an authority score, and flags for likely paywalls and pages that need JavaScript.

Pick the proxy for cheap links at huge volume. Pick an index when search sits inside an agent's loop.

Why the price is structural

Most search APIs are resellers: they hold a commercial agreement with an upstream engine, pay per query under it and add a margin. Their floor is somebody else's wholesale price, and it moves when that contract is renegotiated.

Kaer runs its own crawler and its own index. A query on the web lane costs us capacity, not a fee: the hardware is paid for whether it serves one query or a million, so each additional query costs us very little. That is why the web lane can sit below the resold tier without being a loss leader.

Three lanes, or let it choose

The lane is a parameter on each request, not something you choose at signup, so one key can run a cheap lookup and a deep read seconds apart. Each key handles 60 queries a minute.

Calling it

One endpoint, bearer auth with a kaer_sk_ key that carries the search scope.

curl https://api.kaer.ai/v1/search \
  -H "Authorization: Bearer $KAER_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "query": "post-quantum key exchange in TLS 1.3",
    "lane": "auto",
    "num_results": 20
  }'

The response is { query, lane, served_lane, cost_usd, results }, so every call tells you which lane answered and what it cost. num_results runs from 1 to 25, default 5.

Three statuses worth handling. 402 means the prepaid balance is exhausted: top up and calls resume at once. 429 means the key passed 60 queries a minute and carries a Retry-After. 503 means a lane or the metering was unavailable, and nothing was billed.

Or skip the integration

If what you actually want is for Claude or ChatGPT to be able to search, you do not need any of the above. Kaer speaks the Model Context Protocol: connect it once and your assistant gets search, workflows, mail and calendar as tools, on your account. The walkthrough is here, and the button below creates the connection.

The short version

• $0.50, $1.00 or $3.00 per 1,000, by lane. Up to 25 results at the same price, billed only when a query succeeds, no minimum.
• The web lane is below every verified entry price in the table, including Serper's.
• Serper's largest pack still wins for ten-result queries at volume, and Parallel's Turbo ties our fast lane.
• The model providers' built-in tools cost $10.00 per 1,000 plus tokens.
• The price holds because the index is ours: a query costs us capacity, not someone else's per-query fee.

Frequently asked questions