Brave Search API Coverage in 2026 (Who Resells It, and Who Does Not)

Most “best Brave Search API” lists have the same problem: they name vendors who do not sell the engine. It is an easy mistake to make from the outside, because every one of those vendors sells six or seven other engines and says so on the same page. So this post is a coverage map before it is a price list. I went through the seven largest wholesale SERP vendors in the category one catalogue at a time and wrote down, for each, whether a Brave product exists at all. Six of the seven publish none, and the seventh sells a different product from the one you probably searched for. Checked 2026-09-16.

I read every rate below off the vendor’s own live pricing or catalogue page on 16 September 2026 and linked each one inline, and where a vendor publishes no figure I have said so rather than filled the cell. Nothing here comes from a search snippet, an AI summary or a third-party blog post. I ran no benchmark and I make no claim about anyone’s speed or success rate. Serpent is the API this site sells, and it appears here as one row among several.

Who does not sell a Brave API

Normally an exclusion list is a footnote. Here it is the headline, because the absences are what give this market its shape. If shopping for Brave felt harder than shopping for Google, that is not your imagination — most of the vendors on a normal shortlist simply do not stock it, and none of them make that easy to see. They publish an engine list; Brave is not on it; the page moves on.

Sources: each vendor’s own catalogue, documentation or pricing page, linked in the first column, all read on the same day. An absence on a vendor’s pages is evidence that the vendor does not advertise the product, which is the thing a buyer can act on; it is not proof that no such endpoint exists anywhere inside that company. Where a vendor has told the world what it sells, that list is what a purchase decision has to be built on. Checked 2026-09-16.

One distinction is worth keeping straight, because it changes what you should do about it. “Not in the catalogue” and “not in the price list” are different claims. DataForSEO, Bright Data, SearchApi and Traject Data all enumerate their supported engines explicitly, and Brave is not among them — that is a product decision, stated in public. Scrapingdog’s case is narrower: Brave is simply absent from the credit table that prices every other endpoint it sells, so even if something existed there would be no published number to put in a column. HasData’s catalogue is the widest of the group at 40 APIs and still has no Brave entry. Either way you cannot fill the cell, and inventing the number that fills it is how the comparison posts you already read got written.

It also explains something you may have noticed while shopping. The non-Google engine most of these vendors do carry is DuckDuckGo, and the two are not interchangeable data sets. Brave runs its own crawler and its own index, which is the whole reason a second opinion on a query is worth paying for in the first place. We put numbers on how much two engines really differ in how much search engine results actually overlap, and the engine-by-engine version is Brave versus Google as a SERP data source.

SerpApi’s Brave product is an AI Mode API

SerpApi is the seventh vendor, and it is the subtle one. It does have a Brave product. It is not a Brave web search API.

Its search engine API catalogue listed 112 named APIs on 16 September 2026. Reading it engine by engine: Yahoo has three entries, DuckDuckGo has four, YouTube has four, Google Images has three. Brave has exactly one, and it is the “Brave AI Mode API”. There is no plain Brave Search API in the catalogue, and serpapi.com/brave-search-api — the URL the naming convention of every other engine would predict — returned HTTP 404 on the same day.

That is not a criticism. An AI answer surface is a real product with real buyers, and SerpApi is one of very few vendors selling one for this engine at all. It is a warning about what you are buying. The two things return different shapes and answer different questions:

• A web search product returns the ranked list — position, title, URL, snippet, and whatever blocks the result page carried. That is what rank tracking, competitor monitoring, lead research and retrieval pipelines are built on, because every one of them needs to know which pages won and in what order.
• An AI answer product returns generated prose about the query, with whatever sources it chooses to attach. That is what you want when the deliverable is an answer, and useless when the deliverable is a list of URLs to diff against yesterday’s.

If your shortlist came out of a search for “Brave Search API”, check which of those two you have actually been quoted for before you write a line of parsing code. The same trap exists across the AI-answer category generally, and we mapped which vendors return which answer surfaces in which SERP APIs return AI Overviews.

One more thing that belongs next to SerpApi’s name, because it is a fairness point rather than a sales point: SerpApi publishes one ladder and no per-engine surcharge. None of its pages read on the date above states that one engine costs more than another, and there is no separate rate card per engine. That is not the same as SerpApi promising every engine costs the same, and I have not written it that way. The ladder in the table below is the only rate card SerpApi publishes; we walked it plan by plan in SerpApi pricing explained.

Brave sells its own, and it is the source of record

Before any reseller, the obvious option: Brave sells a Search API itself, directly, with a published rate card. For a large share of readers that is the right answer and the rest of this post is optional reading. It is the source of record for its own index, it is metered rather than bucketed into monthly plans, and it carries a monthly credit that makes small workloads free.

The arithmetic. On the Brave Search API the monthly credit is $5 and the Search plan rate is $5 per 1,000 requests, so the credit covers about 1,000 requests a month before a card is charged. The Answers plan bills a token component on top of its per-request rate, so its true cost depends on answer length and cannot be reduced to one per-1,000 figure — the request rate is the floor, not the bill. The custom plan is quoted, not published, and a quoted price cannot be bounded from outside, so it is marked as unpublished and left alone. Checked 2026-09-16.

The plan names matter for one practical reason: the old Free / Base / Pro ladder is gone. No plan on the page today is named Free, Base or Pro. If you are reading a comparison that describes Brave’s API in those terms — including older pages of ours — it is describing a rate card that no longer exists, and any number attached to it should be re-checked before you budget against it. That is the single most common stale fact about this engine on the web right now.

How I checked this

The question is fixed: can I buy Brave web search results from this vendor, and at what published price? Everything below follows from that.

• A vendor’s own pages are the only source. Search engines were used to find URLs and nothing else. No figure here comes from a search snippet, an AI summary or another blog, and every load-bearing number was read twice.
• An exclusion is a finding and gets named. Six vendors appear in this post with no price at all. Quietly dropping them would leave a tidier table and a dishonest one, because “these are the Brave vendors” and “these are the vendors I found a price for” are very different sentences.
• Entry rate and lowest published rate are separate columns, and the monthly cost of the lowest rate sits next to it. Quoting one vendor’s volume floor against another’s entry price is the commonest dishonesty in this category.
• Only published rates. Brave’s custom plan and SerpApi’s contact-sales tier are both unpriced in public. A price you can negotiate cannot be bounded from outside, so every “lowest published” figure here means exactly that word.
• Different products do not share a row. An AI answer plan and a web search plan are not two prices for one thing, so they are never averaged, compared or ranked against each other.
• Where a figure is my own division rather than a vendor’s printed number, it says so, and the arithmetic is under the table it appears in.

What the realistic options cost

With the exclusions named, the shelf is short. Here is what is actually on it, in the same unit, with each vendor’s entry rate and its lowest published rate kept apart.

Arithmetic and sources. Serpent: the published web category rate card is $0.60 / $0.06 / $0.03 per 1,000 calls at Default, Growth and Scale, with Growth locked by a one-time $100 deposit and Scale by $500; the deposit is spendable balance, not a fee. Quick Search is one unit per request whatever depth you ask for; Deep Search is one unit per page requested, up to 10 pages, so a ten-page Deep call is ten units and the per-1,000 figure is per call, not per page. Brave prints its own rates: “$5 per 1,000 requests” on the Search plan and “$4 per 1,000 requests” plus “$5 per million input/output tokens” on the Answers plan, each with “$5 in free credits every month”; neither is derived. SerpApi prints no per-1,000 rate at all: $25.00 is $25 divided by the Starter plan’s 1,000 searches, and $1.96 is $106,050 divided by Cloud 54M’s 54,000,000 searches, which is $1.9639 exactly. Both SerpApi figures are derived, and both are its general ladder rather than a Brave-specific price. Checked 2026-09-16.

Two things about that table need saying plainly rather than left for a reader to infer.

The Answers row is in the table and out of the ranking. It is there because it is a published Brave rate and leaving it out would hide a real product; it is not ranked against the web search rows because it returns a different thing, and because its token component means the per-request figure is a floor rather than a bill. The same goes for SerpApi’s row: the ladder is real and the arithmetic is right, but what it buys is an AI answer surface.

Read the SerpApi column downward before you quote the small number. $1.96 per 1,000 is a real published rate and it requires the $106,050-a-month Cloud 54M plan, which is 54 million searches. The rate you actually start on is $25.00 per 1,000. That gap between a vendor’s floor and its entry price is the single most-abused number in this category, which is why the two have their own columns here and everywhere else we publish a comparison — the whole-market version is in the SERP API pricing comparison.

What to check before you call it Brave coverage

“Supports Brave” on a feature grid is one tick box covering at least five separate questions. This is the part of a provider comparison almost nobody does, and it is the part that decides whether the cheap option was cheap. Ask every candidate the following, and note that I am answering for our own API from what we publish rather than from a benchmark — I did not run paid calls against anyone else for this post.

• Which Brave surfaces, not just “Brave”? Web results are one surface; news, images and video are three more, and a vendor that has one may not have the others. Ours are web (a Quick and a Deep variant), news, images and videos, each documented on the Brave Search API page with its own published category on the rate card.
• How many rows does one call return? This is the number that turns a price per call into a price per result, and most vendors never print it. On our side num accepts up to 100 on web search and up to 50 on news, and Brave’s own result pages carry roughly ten to twenty organic rows each, so a bigger num is a ceiling rather than a promise.
• Country and language targeting. Ours takes 112 country codes — GET /api/countries returns the current list — and a two-letter ISO 639-1 language code. Language coverage is at its broadest on this engine, which is one of the genuinely good reasons to query it.
• Freshness, and its floor. A real limitation worth knowing before you design around it: Brave’s narrowest time window is the past day, so asking for the past hour narrows no further than asking for the past day on this engine. If you need hour-level recency, this is not the engine for that query.
• What a missing block looks like. The question that decides whether a multi-engine parser is pleasant or miserable. On our API every results key is present on every call, so a block the page did not carry arrives as null or [] rather than as a missing key, and a call that came back short says so in meta.partialResults instead of quietly looking complete. A call that returns nothing at all is refunded in full.
• What the engine does not have. Two of ours, stated plainly: pixel_position is not available on this engine, so if you need on-page geometry you need a different one; and on image search this engine accepts a language but not the size, type, colour, aspect, licence or time filters the others take. A vendor that claims every parameter works on every engine has not tested it.

The reason to ask all six is that the cheapest row on a price table is not the cheapest option if it returns half the rows or drops the field your pipeline keys on. If you are merging several providers into one schema, the shape questions matter more than the rate, and the pattern for doing it is in normalising SERP JSON across providers.

Where the other options beat us

A comparison that always concludes “buy ours” is worth nothing to you and nothing to a search engine. Here is where we lose, stated as plainly as the rest.

Small workloads, on price, outright. On the Brave Search API, $5 in free credits every month against a published $5 per 1,000 requests is about a thousand requests a month at zero cost, and no per-call rate undercuts free. If your entire job is a daily check on a few dozen queries, buy nothing: go to brave.com/search/api, take the credit, and skip this whole market.

Authority, always. Brave is the source of record for its own index. If you are publishing research about what Brave returns, or your contract requires a first-party data source, that is a reason that no price on this page outweighs. The same applies to any dispute about whether a result was really there.

An AI answer surface for this engine. SerpApi sells a Brave AI Mode API and Brave sells its Answers plan. If generated prose is the deliverable, those are the two products built for it, and a web search API is the wrong tool no matter what it costs.

Cached searches. Every SerpApi engine page carries the same line: “Cached searches are free, and are not counted towards your searches per month.” On a workload with heavy query repetition that is a real discount which no rate card captures. On any provider that bills every call, the equivalent saving is yours to build, and the pattern is in building a SERP cache to cut your API bill.

Free tiers for evaluation. SerpApi gives 250 searches a month, every month, which is more evaluation headroom than our free calls to start. Ours is enough to read the JSON and decide whether you like the field names; it is not a pilot.

Which option fits which job

• Under about a thousand requests a month. Brave’s own API, on the monthly credit. It costs nothing, it is first-party, and nothing here beats free.
• Steady commercial volume on web results. The cheapest published route in this comparison is our Brave web search endpoint on the web category rate, which needs no plan and no monthly commitment. The published alternative is the Brave Search API at $5 per 1,000 requests, bought direct: simple, first-party, and roughly eight times our entry rate.
• A generated answer rather than a ranked list. SerpApi’s Brave AI Mode API, or the Answers plan bought direct. Do not buy a web search API for this job.
• Brave alongside other engines on one key. On our API the engine is a query parameter, so the same endpoint and the same key return Brave, Google, Bing, Yahoo or DuckDuckGo with the same response shape; the multi-engine aggregator walkthrough shows the pattern, and how much search engine results actually overlap is the evidence for whether the second engine is worth querying at all.
• An AI agent or RAG pipeline that needs a non-Google opinion. Since the Bing Search API was withdrawn, the independent-index shelf is short — the replacements are catalogued in Bing Search API alternatives and the agent-specific trade-offs in the best web search APIs for AI agents and RAG.
• Zero budget and engineering time to spend. Building it yourself is a real option and an underrated cost. The honest version of that trade-off is written up in scraping Brave Search for free, and the walkthrough is scraping Brave search results. Read the maintenance section before you commit a sprint to it.
• Still deciding whether Brave is the right index for your question. That is a data question rather than a price one, and Brave versus Google as a SERP data source is where it is answered. There is no point optimising the price of data you did not need.

Whichever you pick, do the same two checks before you commit a budget. Run one real query and count the organic rows you got against the number you assumed, because almost nobody publishes it. Then run one query that returns nothing at all — a nonsense string is fine — and look at what comes back, because an empty result and a broken request should never wear the same status code. Those two calls will tell you more than any comparison table, this one included.

FAQ

Which SERP API vendors sell Brave search data in 2026?

Fewer than the size of the category suggests. Of the seven largest wholesale SERP vendors, six publish no Brave product at all: DataForSEO, Bright Data, SearchApi, HasData, Scrapingdog and Traject Data. Each of those was checked against that vendor's own engine list, product catalogue or credit table on 16 September 2026, and Brave appears on none of them. The seventh, SerpApi, lists a Brave AI Mode API and no plain Brave web search product. Brave itself sells a Search API directly, and a small number of general search APIs carry Brave as one engine among several. Checked 2026-09-16.

Does SerpApi have a Brave Search API?

It has a Brave AI Mode API, which is not the same product. SerpApi's search engine catalogue listed 112 named APIs on 16 September 2026, and the only Brave entry among them is the Brave AI Mode API. The URL you would guess for the other one, serpapi.com/brave-search-api, returned HTTP 404 the same day. The distinction matters at integration time: an AI answer surface hands back generated text, while a web search product hands back a ranked list of links with titles, URLs and snippets. If your code parses organic rows, the AI Mode product is not a drop-in for it. Checked 2026-09-16.

How much does Brave's own Search API cost?

It publishes two metered plans and one custom plan. On the Brave Search API the Search plan is $5 per 1,000 requests, with 50 queries per second. The Brave Search API Answers plan, which returns a synthesised answer rather than a list of links, is $4 per 1,000 requests plus $5 per million input and output tokens, with 2 queries per second. Both metered plans include $5 in free credits every month. The third plan covers custom terms, capacity or endpoints and carries no published rate at all, so nobody outside can bound what it costs. Checked 2026-09-16.

Is there still a free Brave Search API?

Not as a standing free tier, but not nothing either. Both metered plans on the Brave Search API include $5 in free credits every month, and at the Search plan's published $5 per 1,000 requests that credit is worth about 1,000 requests a month before a card is charged. No plan on the page today is named Free, Base or Pro; write-ups that still describe that ladder, ours among them, are out of date. If a thousand requests a month covers your job, the cheapest way to get Brave data is Brave's own API, and nothing on this page undercuts it. Checked 2026-09-16.

What is the difference between a Brave web search API and a Brave AI answer API?

They answer different questions and they return different shapes. A web search product returns the ranked list: position, title, URL, snippet, and whatever blocks the page carried, which is what rank tracking, competitor monitoring and retrieval pipelines are built on. An AI answer product returns generated prose about the query, with whatever sources it chooses to attach, which is what you want when the deliverable is an answer rather than a list. Both are sold under the word Brave in 2026, and a buyer who searches for one can easily end up reading the rate card for the other. Checked 2026-09-16.

Which option should I choose for Brave data?

It depends on volume and on what you need back. Under about a thousand requests a month, the monthly credit covers you and Brave's own API costs nothing, and it is the source of record for its own index. Above that, the Brave Search API Search plan is the rate to compare against, at $5 per 1,000 requests. Serpent carries Brave as one of five web engines on the same endpoint and the same key, billed on its published web category rather than on a Brave-specific price. If what you want is generated prose rather than a ranked list, SerpApi's Brave AI Mode API and the Answers plan are the two products that sell it. Checked 2026-09-16.

Related reading, if you came here from the product side of this engine rather than the web side: the same exercise on Brave product results is in the Brave shopping API comparison.