For architects
Web search APIs for AI agents — the 2026 landscape
We sell one of the products on this page, so treat the recommendations accordingly. The facts, however, are all sourced: every figure below was read from the vendor's own pricing page or funding announcement, each one linked with the date we read it. Where a vendor does not publish a comparable rate, we say so rather than estimating.
Verified 7 October 2026·All figures from primary sources
- providers tracked
- 20
- run their own index
- 10
- publish a comparable per-1,000 rate
- 13
- retired, closing or acquired
- 3
each with dated sources
Price per 1,000 queries
Comparing prices across this market is genuinely hard: some vendors bill per request, some per credit with different credit costs per endpoint, and some per token on top of a per-request search fee. Every figure here is normalised to the same thing — one plain web search returning about ten results, with no extraction and no summarisation.
- unlob$0.40
- DataForSEO$0.60
- Parallel Search$1
- Serper$1
- Perplexity$1
- Bright Data$1.50
- Exa$4
- Brave Search API$5
- Linkup$5
- Google Custom Search JSON API$5
- Tavily$8
- Grounding with Google Search$14
- Grounding with Bing Search$14
- SerpAPI$25
| Provider | Own index? | $ / 1,000 searches | Source |
|---|---|---|---|
Priced: Starter plan. $25 per 1,000 searches on the Starter plan ($25 a month for 1,000), falling to about $1.96 per 1,000 on the largest Cloud plan. | No Scrapes Google, Bing, Baidu and others | $25 | SerpApi pricing |
Priced: search queries on Gemini 3.x. $14 per 1,000 search queries on Gemini 3.x models after 5,000 free search requests a month (paid tier, shared across Gemini 3.x), plus model token charges. One prompt can trigger several billed queries. Gemini 2.5 models keep the older scheme: $35 per 1,000 grounded prompts after 1,500 free a day. | Yes Google's index | $14 | Gemini API pricing |
Priced: Grounding with Bing Search transactions (one per tool call). $14 per 1,000 transactions, counted per tool call within an agent run, with model costs billed separately. Available only inside Azure AI Foundry Agent Service or as a web knowledge source in Azure AI Search, and the raw results are not returned to the developer. | Yes Bing index, via Azure AI Foundry | $14 | Grounding with Bing Search |
TavilyAcquired Priced: basic search, pay as you go. Credit-based at $0.008 per credit pay-as-you-go. A basic search is 1 credit (~$8 per 1,000), an advanced search 2 credits (~$16 per 1,000), and a research request 4 to 250 credits. Fast and ultra-fast searches are also 1 credit; monthly plans bring the effective rate down to $5 per 1,000 at 100,000 credits ($500 a month). | No No publicly documented independent web index | $8 | Tavily pricing |
Priced: Search plan. Search is $5 per 1,000 requests. The Answers plan is $4 per 1,000 plus $5 per million tokens. | Yes 40B+ pages, no Google or Bing dependency | $5 | Brave Search API pricing |
Priced: standard-depth search. $5 per 1,000 standard-depth searches returning results ($0.005 a call); $6 per 1,000 with a sourced answer, and $50 per 1,000 at deep depth. | Partial Own retrieval layer over partner and web sources | $5 | Linkup pricing |
Priced: paid queries (existing customers only). $5 per 1,000 queries after 100 free a day, up to 10,000 queries a day — for existing customers only. The API is closed to new customers and is discontinued on 1 January 2027. | Yes Google's index, via Programmable Search Engine | $5 | Custom Search JSON API pricing |
Priced: Instant search. Instant search is $4 per 1,000 requests (up to 10 results); Fast and Auto are $7 per 1,000, and the Deep modes $12 to $15. Results beyond 10 add $1 per 1,000, contents $1 per 1,000 pages per content type, and AI summaries $1 per 1,000 pages — so a realistic RAG call costs more than the headline. | Yes Proprietary AI-native semantic index | $4 | Exa pricing |
Priced: SERP API, pay as you go. The SERP API returns scraped search-engine results pages at $1.50 per 1,000 requests pay as you go; the $499-a-month Scale plan includes 380,000 requests, then $1.30 per 1,000. Proxy and browser products are priced per GB. | No Proxy, unblocking and scraping infrastructure | $1.50 | Bright Data SERP API pricing |
Priced: Turbo processor. The Turbo and Fast search processors are $1 per 1,000; Basic and Advanced are both $5 per 1,000. The Task API, which is what most agent workloads actually use, runs $5 to $2,400 per 1,000 depending on processor. | Yes Proprietary web index and retrieval infrastructure | $1 | Parallel pricing |
Priced: Starter credit pack. $1 per 1,000 on the Starter credit pack, falling to $0.30 per 1,000 on the largest pack. Prepaid credits, valid for six months. | No Google SERP wrapper | $1 | Serper pricing |
Priced: Search API with Fast Search. The Search API is $5 per 1,000 requests, or $1 with Fast Search, with no token charges (up to 20 results, 10 by default). Sonar Chat Completions support ended on 27 September 2026 and requests are moving to the Agent API; Sonar charged $5 to $12 per 1,000 by search context size plus $1/$1 per million tokens, and Sonar Pro $6 to $14 plus $3/$15. | Yes Perplexity index | $1 | Perplexity API pricing |
Priced: Standard queue (asynchronous). $0.60 per 1,000 SERPs (10 results each) in the Standard queue, which is asynchronous: you post a task and collect the result later. Priority queue is $1.20 per 1,000 and real-time Live mode $2. Pay as you go, with a $50 minimum payment. | No SERP data collection at scale | $0.60 | DataForSEO SERP API pricing |
$29 per month for 50,000 credits, and a search is one credit. | Yes Our own crawl and index | $0.58 | This page |
$199 per month for 500,000 credits, and a search is one credit. | Yes Our own crawl and index | $0.40 | This page |
How the market got here
Two years that changed what buying a search API means. Every entry is a public announcement or a vendor's own notice.
Microsoft retired the Bing Web Search API
The whole family — Web, Image, News, Video, Entity, Autosuggest, Spell Check and Custom Search. New resource creation was quietly blocked in early 2025, months before Microsoft’s retirement notice of 15 May 2025, so the retirement was already irreversible by the time it was announced.
Brave restructured its Search API plans around a monthly credit
Free use is now a $5 monthly credit per plan against paid usage, conditional on attributing Brave. The index is unchanged; what changed is that a hobby project is now metered.
Nebius Group acquired Tavily for $275 million
Tavily had raised $25M across a seed and a $20M Series A led by Insight Partners. Its search still runs on an upstream provider rather than an index it owns.
Parallel Web Systems reached a $2.0B valuation
$100M Series B led by Sequoia, bringing total raised to $230M. Founded by Parag Agrawal, and running its own index.
Exa Labs raised $250M at a $2.2B valuation
Led by Andreessen Horowitz, after an $85M Series B at $700M in September 2025 — roughly tripling the valuation in eight months.
Two things happened at once. Owning an index stopped being a differentiator — Exa, Parallel and Brave all run their own, and the two largest are capitalised at over two billion dollars each. And the cheap on-ramps closed: Bing’s API is gone, Brave’s free use is now a metered monthly credit, and Tavily now answers to an acquirer. The interesting question in 2026 is no longer who owns an index. It is what a given index will tell you about itself, and what it charges.
Capability matrix
Four questions that decide most architectures: does the provider control its own index, can an agent call it over MCP, can it traverse relationships between documents, and will it tell you what it left out.
| Provider | Own index | MCP | Graph traversal | Coverage transparency | Free tier |
|---|---|---|---|---|---|
| unlob | Yes | Yes | Yes — 6 tools | Yes — why_not and per-call receipts | 10,000 credits/mo |
| Exa | Yes | Yes | No | No | $10 in credits at signup (up to 2,500 Instant searches), reset to $10 on the first of every month |
| Parallel Search | Yes | Yes | No | No | 5,000 requests per month |
| Brave Search API | Yes | Yes | No | No | $5 in free credits every month (about 1,000 Search requests), conditional on attributing the Brave Search API; plans restructured 12 February 2026 |
| Mojeek | Yes | No | No | No | Free trial with limited queries, on request |
| TavilyAcquired | No | Yes | No | No | 1,000 credits per month, no card required |
| Linkup | Partial | Yes | No | No | $20 of credit, topped back up to $20 each month for eligible accounts (about 4,000 standard searches) |
| Valyu | Partial | Yes | No | No | $10 in free credits ($20 with a work email) |
| SerpAPI | No | Yes | No | No | 250 searches per month |
| Serper | No | Yes | No | No | 2,500 free queries on signup |
| DataForSEO | No | No | No | No | $1 trial credit, no deposit required |
| Google Custom Search JSON APIClosing | Yes | No | No | No | 100 queries per day (existing customers only) |
| Grounding with Google Search | Yes | No | No | No | 5,000 search requests per month on the paid tier (Gemini 3.x); on the free tier only Gemini 2.5 Flash and Flash-Lite, up to 500 requests a day |
| Perplexity | Yes | Yes | No | No | None stated; API credits are prepaid |
| Grounding with Bing Search | Yes | No | No | No | None — paid or pay-as-you-go Azure subscriptions only; free-credit subscriptions are not eligible |
| Bing Web Search APIRetired | Yes | No | No | No | None — retired |
| OpenAI web search tool | Partial | No | No | No | None separate from API credit |
| Firecrawl | No | Yes | No | No | 1,000 credits per month, no card required |
| Jina Reader | No | No | No | No | 10 million tokens for non-commercial use; Reader works without a key at 20 requests / min, Search needs a key |
| Diffbot | Yes | No | No | No | 10,000 credits per month, no card required |
| Bright Data | No | Yes | No | No | 5,000 SERP API requests per month (and 5,000 Web MCP requests), no card required |
"Graph traversal" means the API can return relationships between documents in the open web index — not that you can build a graph over your own documents with it. "Coverage transparency" means the API will tell you why a specific URL is absent.
Who is funded, and by whom
Worth knowing before you pick a dependency. Heavy funding means the provider will still be there next year; it also means the pricing you sign up on is not necessarily the pricing that has to clear a $2 billion valuation later.
| Company | Total raised | Last round | Valuation | Lead investors |
|---|---|---|---|---|
| Exa Exa Labs | $335M+ | $250M (2026) May 2026 | $2.2B | Andreessen Horowitz, Benchmark, Lightspeed |
| Parallel Search Parallel Web Systems | $230M | $100M Series B April 2026 | $2.0B | Sequoia Capital, Kleiner Perkins, Index Ventures |
| Tavily Nebius Group | $25M | $20M Series A August 2025 | — | Insight Partners, Alpha Wave Global |
| Linkup Linkup | $10M+ | $10M seed February 2026 | — | Gradient, Elaia, Seedcamp |
| Firecrawl Firecrawl | $16.2M | $14.5M Series A August 2025 | — | Nexus Venture Partners, Y Combinator, Tobias Lütke |
unlob is not on this table: we are self-funded and not raising. That is a fact about us, not an argument — it means our pricing has no valuation to grow into, and it means we are smaller than everyone above.
The five kinds of provider
These are usually compared as one market. They are not one market — the failure modes are completely different, and picking across categories by price is how teams end up rebuilding.
Own-index search APIs
These vendors crawl and index the web themselves. Nobody upstream can deprecate them, reprice them or rate-limit them.
Neural, embeddings-first search over a proprietary index, sold alongside content extraction, answers and agentic research products.
Source: Exa pricing, Exa rate limits, Exa Labs raises $250M at $2.2B valuation, Announcing our Series B
A suite of web search, extraction and research APIs on a proprietary index, from the former Twitter CEO. The closest competitor to unlob by thesis.
Source: Parallel pricing, Parallel Web Systems hits $2B valuation
A genuinely independent index of 40B+ pages, sold as a conventional search API with an LLM-context option. Long the default independent choice.
Free developer tier removed February 2026
Source: Brave Search API pricing, The most powerful search API for AI (12 February 2026)
A UK independent crawler and index, running since 2004. Small by comparison, but genuinely its own — and the oldest independent index still standing.
Source: Mojeek
Raw web search over its own index through the Search API, and search with synthesis through the Agent API, which replaced Sonar Chat Completions in September 2026.
Source: Perplexity API pricing, Perplexity Fast Search, Migrating from Sonar to the Agent API
Platform grounding
Search bundled into a model platform. Nothing to integrate if you are already there, and no control over retrieval if you are not.
Google's first-party grounding for Gemini. The best index in the world, billed per search query on top of tokens, and only usable from Gemini.
Source: Gemini API pricing
Microsoft's designated successor to the retired Bing Search API, available only inside Azure AI Foundry. A platform commitment, not an API swap.
For a decade the default independent web search API. Retired on 11 August 2025, taking Web, Image, News, Video, Entity and Custom Search with it.
Retired 11 August 2025
Source: Bing Search API retirement
The built-in search tool in the Responses API. Zero integration work if you are already on OpenAI, and no control over retrieval whatsoever.
Source: OpenAI API pricing
Retrieval layers
These sit on top of other search providers and add cleaning, ranking and citation. Convenient, with someone else’s index underneath.
An LLM-shaped retrieval layer returning cleaned, ranked, cited chunks rather than raw SERPs. Acquired by Nebius in February 2026.
Acquired by Nebius Group for $275M, February 2026
Source: Tavily pricing, Tavily API credits, Nebius announces agreement to acquire Tavily (10 February 2026), Tavily raises $25M, Tavily company profile
A French grounding API returning structured, citation-carrying content for LLMs, with a sub-second "fast" tier and native parallel search.
Source: Linkup pricing, Linkup pricing documentation, Linkup raises $10M seed
A London deep-search API spanning web plus licensed proprietary sources, with content attribution and pay-per-use monetisation for publishers.
Source: Valyu pricing, Valyu pricing documentation
SERP resellers
These return results from Google or Bing. Cheap and familiar, but the product exists at the pleasure of a search engine that does not want it to.
The mature multi-engine SERP API. Returns real Google, Bing and Baidu result pages with all their features — at SERP-scraping prices.
Source: SerpApi pricing
The cheapest way to get raw Google results into an agent. Fast, minimal, and entirely dependent on Google continuing to be scrapeable.
Source: Serper pricing
Industrial SERP infrastructure built for SEO tooling rather than agents. The cheapest per query in the field, with the lowest-level API to match.
Source: DataForSEO SERP API pricing
Google's own programmable search endpoint. Real Google results, a hard 10,000/day ceiling, and terms written for site search rather than agents.
Closed to new customers; Google discontinues it on 1 January 2027 and points site search to Vertex AI Search
Source: Custom Search JSON API pricing
Crawl & extraction
These turn a URL into clean text. Essential in most stacks, but they need something to hand them the URLs.
Crawl, scrape and clean any URL into markdown for an LLM. A complement to a search API rather than a replacement — it needs URLs to start from.
Source: Firecrawl pricing, Firecrawl search, Firecrawl Series A
Prefix any URL and get clean markdown back. The lowest-friction extraction tool there is, and priced per token rather than per page.
Source: Jina Reader, Jina API pricing
One of the very few independent commercial web crawls, feeding a structured Knowledge Graph. Entity-shaped rather than passage-shaped.
Source: Diffbot pricing, Diffbot Web Search API
Proxy and unblocking infrastructure with a scraping suite and an agent-facing MCP server. Built for volume collection, priced accordingly.
Source: Bright Data SERP API pricing, Bright Data Web MCP pricing
Take the data
This table is published as JSON, with the source URL and read date attached to every figure. Cite it, check it, or use it to build your own comparison — we would rather the numbers were right than exclusive.
/api/landscape.jsonFrequently asked questions
Which web search APIs actually own their index?
Exa, Parallel, Brave, Mojeek, Diffbot and unlob crawl and index the web themselves. Google Custom Search, Gemini grounding, Perplexity and the retired Bing API serve a large first-party index but only on their own terms. SerpAPI, Serper and DataForSEO resell Google or Bing results; Tavily and Linkup add a retrieval layer over other providers; Firecrawl and Jina Reader extract from URLs you already have.
What happened to the Bing Web Search API?
Microsoft retired it on 11 August 2025, taking Web, Image, News, Video, Entity, Autosuggest, Spell Check and Custom Search with it. New resource creation was quietly blocked in early 2025, months before the 15 May 2025 retirement notice. The official migration path is Grounding with Bing Search inside Azure AI Foundry, which is a platform commitment rather than a drop-in replacement.
Does the Brave Search API still have a free tier?
Not the perpetual one. Brave retired its long-running free developer tier in February 2026 and replaced it with a $5 monthly credit against paid usage. The index itself — 40 billion pages and genuinely independent of Google and Bing — is unchanged.
Is Tavily still independent?
No. Nebius Group agreed to acquire Tavily for $275 million in February 2026, after Tavily had raised $25 million across a seed and a $20 million Series A led by Insight Partners. The API continues to operate as part of Nebius.
How much money has gone into this category?
Roughly $600 million in disclosed funding plus a $275 million exit, in about eighteen months. Exa raised $250 million in May 2026 at a $2.2 billion valuation; Parallel Web Systems raised $230 million across two rounds, reaching $2.0 billion in April 2026; Firecrawl raised a $14.5 million Series A; Linkup raised a $10 million seed.
Which is cheapest per 1,000 queries?
Among providers that publish a flat rate and run their own index, unlob is cheapest at $0.40 per 1,000 searches on the Scale plan (a search is one credit), ahead of Parallel's Turbo tier at $1, Brave at $5 and Exa at $4. DataForSEO is cheaper still at roughly $0.60 to $1, and Serper reaches $0.30 at volume, but both are SERP resellers dependent on Google.
Does any search API expose a knowledge graph over the open web?
Diffbot maintains a Knowledge Graph from its own crawl, and unlob exposes a coverage graph with six traversal tools — related, corroborate, authorities, dossier, path, assemble_context. Beyond those, graph retrieval in this market is generally something you build yourself over your own documents with GraphRAG or Neo4j, rather than something a web search API provides.
The cheapest own-index option on this page
$0.40 per 1,000 searches, a coverage graph nobody else offers, and an answer when you ask why a URL is missing. Start on the free tier and check the claims yourself.
Evaluating for a team or at volume? Send a workload brief or write to hello@unlob.com — a person replies.