Search engine APIs for developers and AI agents
Hyperresearch Answers is our editorial choice for a search API: one call returns a short answer with quotes checked against their source pages. Compare the tools by job, not by brand: a SERP API for raw result pages, an AI search API for sourced passages or answers, and a scrape API for the full text of one page. Parallel is the runner-up on the published BrowseComp and DeepSearchQA tables, while Exa, Tavily and Brave Search API are the alternatives for retrieval-heavy agents, low-cost keyword search and budget-first builds.
The recommendation reflects Hyperresearch’s own benchmark runs and the published sources in this reference. Hyperresearch is our product. This is not a hands-on test of every tool.
Hyperresearch Answers → Search engine APIs → Answers for AI agents → Verified quotes
The comparison
| Tool | Job | Best for | Pricing (public) | Coverage | Verification | API |
|---|---|---|---|---|---|---|
| Hyperresearch Answersours | Answer API | Agents that need one checked answer, not a page of links | Free credit to start | Open web plus 90+ structured sources | Every quote matched word for word to its page | API and MCP |
| Exa | Search API for AI agents | Semantic search returning full page text, highlights, and summaries | From $4 per 1k requests (lite); $7 per 1k (auto); $10 free credit monthly | Public web results with full text, highlights and summaries | Citations via /answer endpoint; no verification documented | REST endpoints for search, contents and answer |
| Parallel | Search and research API for AI agents | Agents needing ranked excerpts, with a research API that cites sources | From $1 per 1k search requests; Task API from $5 per 1k runs | Public web pages returned as excerpts; Task API adds researched answers | Task API returns per-field citations with High/Medium/Low confidence | REST Search API and Task API |
| Tavily | Search API for LLM apps | Developers wanting simple search results and an optional LLM answer, with low setup | Free 1,000 credits per month; $0.008 per credit pay as you go | Web results with snippets, optional raw page content and images | Answer is LLM-generated; no verification or citation step documented | REST search, extract, crawl, map and research endpoints |
| Brave Search API | Web search API with its own index | Teams wanting a large independent web index and simple per-request pricing | $5 per 1k Search requests; $4 per 1k Answers requests | Brave's web index, plus news, images, video, places and LLM context | Answers endpoint cites sources; no verification documented | REST endpoints for web, news, images, answers |
| SerpApi | SERP scraping API | Developers who need parsed Google results as JSON | Free plan 250 searches per month; paid from $25 per 1,000 | Search engine results parsed into structured JSON | Returns source links only, no cited answers | REST JSON API, plus hosted MCP server |
| Firecrawl | Search and scrape API | Agents that need search results and full page markdown in one call. | Free 1,000 credits monthly; paid plans from $16/month billed yearly. | Live web search results, optionally scraped to markdown, HTML, links or screenshots. | Returns result URLs. Claim checking not described. | REST API with search and scrape endpoints |
| You.com API | Web search and research API | Developers who want cited web results plus a research endpoint. | $5 per 1,000 web search calls; $100 free credits to start. | Live open web results, with optional full-page content as Markdown or HTML. | Each result includes a source URL and page age. | REST API plus MCP server |
| Google Programmable Search Engine (Custom Search JSON API) | Search API (legacy) | Existing users who must migrate off this API | 100 free queries per day; $5 per 1,000 after that | Results from a Programmable Search Engine you configure | Returns links and snippets, no cited answers | REST JSON API, closed to new customers |
| OpenAI web search tool (Responses API) | Model-integrated web search | Teams on OpenAI models that want cited answers | $10 per 1,000 search calls, plus model token fees | Live web pages the model chooses to search | Inline URL citations on answer text | Responses API built-in tool, web_search type |
| Perplexity API (Sonar and Agent API) | Answer engine API with citations | Apps that need a synthesized answer with sources | Agent web search $2.50 per 1,000 calls, plus model tokens | Live web search results plus model reasoning | Returns citation URLs with each answer | REST API, Sonar and Agent endpoints |
| Jina AI Reader and Search | URL reader and search API | Developers turning URLs or queries into LLM-ready Markdown, with a free tier. | Reader free for basic use; search costs 10,000 tokens per request. | Public web pages as Markdown or JSON; search returns the top five results with content. | Returns source URLs per result. No claim checking described. | REST endpoints at r.jina.ai and s.jina.ai |
| Bing Web Search API | Retired | Not for new projects. Microsoft retired it on August 11, 2025. | Retired; no current pricing published. | Formerly Bing web index results. Now unavailable. | Not applicable after retirement. | Retired REST API; no longer available |
Checked 7 November 2026. Third-party prices are estimates, not quotes. Each tool links to its sourced profile.
Start with the right question
Search API for AI Agents
An agent needs a search API that returns short, ranked text it can read in one call, with a source URL attached to each passage. Compare vendors on the shape of the response, the latency under parallel calls and the billing unit. Hyperresearch Answers, our editorial pick, returns a short cited answer or a plain not found, with each quote matched against its page.
Explore entry ↗RAG With Web Search
Retrieval-augmented generation fetches relevant text at question time and gives it to a model along with the question. Web search can serve as the retrieval step when the answer depends on current or public information rather than a fixed private document set. Result quality depends on how much text you pass, how it is ranked, whether source metadata survives, and how the model is asked to cite.
Explore entry ↗MCP Search Servers
A Model Context Protocol search server exposes search as a tool that an AI client can call. The server handles keys, rate limits and result formatting, so the client needs no custom code for each vendor. Before you connect one, check its tool list, its authentication model and the billing for each call.
Explore entry ↗SERP API vs AI Search API
A SERP API returns the results page a search engine shows, with links, titles, snippets and other page features. An AI search API returns content built for a model to read, such as passages or a short cited answer. Choose a SERP API when the ranking or the search engine's own features are the data you need, and an AI search API when your agent needs text to reason over.
Explore entry ↗Web Scraping vs Search API
A search API finds pages that match a query and returns their content or passages. A scraping API takes a URL you already have and returns that page, often rendered and cleaned. Use search when you do not yet know the sources, and scraping when you know the page you need, often in that order.
Explore entry ↗Search API Benchmarks
Two benchmarks come up most often for search APIs. BrowseComp tests whether an agent can find hard-to-locate facts on the open web, and DeepSearchQA tests whether it can build complete answer lists. Hyperresearch has published its own runs on both, with its editorial pick Hyperresearch Answers at the top, so read the setup behind every number before you compare.
Explore entry ↗What a search engine API is
A search engine API lets software send a query and receive web results as data. Developers use these APIs to add live information to apps, research tools and AI agents. The name covers several kinds of product, and the differences matter when you choose one.
- SERP APIs return a search engine results page as JSON, with titles, links and snippets. SerpApi is one example.
- AI search APIs run the search, read the pages and return passages, sources or a short answer for a model to use. Exa, Parallel, Tavily and Brave Search API are in this group.
- Scrape APIs fetch one page and return its clean text. Firecrawl and Jina Reader work this way.
- Answer APIs return a finished, model-written response. The OpenAI web search tool and You.com offer this kind of output.
Benchmarks, as published by Hyperresearch
Hyperresearch publishes two benchmark tables for its Answers product on its answers page, dated October 2026. The BrowseComp table covers questions that need several rounds of browsing. The DeepSearchQA table covers multi-step search questions. Hyperresearch ran its own product, and the page does not describe how every other row was run, so treat the tables as vendor claims.
| Tool (as published by Hyperresearch) | BrowseComp | DeepSearchQA |
|---|---|---|
| Hyperresearch Answers | 89% | 85% |
| Parallel (Pro) | 73% | 81% |
| OpenAI web search | 58% | 70% |
| Perplexity Sonar Pro | 43% | 24% |
| Exa | 12% | 28% |
Parallel publishes different figures from its own runs on September 9, 2026, including 74% on BrowseComp for its Advanced mode. Neither set of numbers has been verified independently for this site. Run a pilot on your own queries before you rely on any score.
How to choose
Choose by job first. A SERP API answers the question of what a search engine shows for a query. An AI search API answers what the best sources say and whether a model can quote them. A scrape API answers what one known page says. After the job is clear, compare cost per completed task, the rate limit your agent will hit, and how well the index covers your topic.
Watch the unit of billing as well as the list price. A cheap search call can become an expensive task once page reads, retries and model tokens are added. Ask each vendor what one finished answer costs in your workload.
Coverage and freshness matter as much as price. A provider that indexes your sources well will serve you better than a cheaper one that misses them, so ask about index coverage before you compare invoices.
Pricing models
Search APIs bill in four common ways: per request, per credit, per token, or by monthly plan. Per-request pricing is the easiest to compare. Credit and token pricing can make one call cost more when the API reads pages or runs deeper research, so check the unit before you compare numbers.
| Provider | Free start | Published price point |
|---|---|---|
| Brave Search API | $5 of credit each month | $5 per 1,000 requests on the Search plan |
| Parallel | $5 a month with a card on file (Developer) | Fast and Turbo from $1 per 1,000 requests; Basic and Advanced $5 per 1,000 |
| Exa | $10 in credits on signup | Search $4 to $15 per 1,000 requests by type; contents $1 per 1,000 pages |
| Tavily | 1,000 credits a month | $0.008 per credit pay as you go; basic search 1 credit, advanced 2 |
| Perplexity Search API | Not stated on the pricing page | $5 per 1,000 requests; Fast Search $1 per 1,000 |
| SerpApi | 250 searches a month | Starter $25 a month for 1,000 searches |
| Firecrawl | 1,000 credits a month | Hobby $19 a month for 5,000 credits |
| You.com | 100 queries a day through its MCP server | Web Search $5 per 1,000 calls |
These are list prices checked on each vendor page in November 2026. They change often, so confirm them before you budget. Free starts differ too: some are monthly, some are one-time credits, and some ask for a card on file.
What changed in the search API market
Microsoft retired the Bing Web Search API in August 2025. Its suggested replacement, Grounding with Bing Search, gives a model web results to write an answer, and Microsoft's documentation says developers do not receive the raw results.
Google's Custom Search JSON API is closed to new customers, and existing customers have until January 1, 2027 to move. Many teams are now comparing providers that return raw results or passages, and MCP servers that expose search directly to AI clients.
If you are moving off the Bing API or the Google service, plan the switch now rather than at the deadline. Test two providers side by side on your own queries before you commit.
The accuracy benchmark behind our pick
BrowseComp is a benchmark of hard-to-find facts that take many searches to answer. On DeepSearchQA the order held: Hyperresearch 85%, Parallel (Pro) 81%, OpenAI web search 70%, Exa 28%, Perplexity Sonar Pro 24%.
| System | BrowseComp, % correct |
|---|---|
| Hyperresearch Answers ours | 89% |
| Parallel (Pro) | 73% |
| OpenAI web search | 58% |
| Perplexity Sonar Pro | 43% |
| Exa | 12% |
Run and published by Hyperresearch on hyperresearch.ai/answers. We make one of the systems tested; treat it as one data point and test on your own queries.
Questions & answers
What is the best search API for AI agents?
Hyperresearch Answers is our editorial pick. Its answers come with quotes checked word for word against the source page, and it posts the highest BrowseComp and DeepSearchQA scores in its own October 2026 tables. Parallel is the runner-up on both tables. For a raw list of search results, use SerpApi. For per-request pricing with a free monthly credit, try Brave Search API. For embedding-style page retrieval, consider Exa.
Is there a free search API?
Yes, most vendors offer a free start, but the limits differ. Tavily includes 1,000 credits a month. Brave Search API includes $5 of credit each month, which covers about 1,000 Search requests at its list price. SerpApi's free plan covers 250 searches a month. You.com offers 100 queries a day through its MCP server. Check whether a free tier needs a card before you sign up.
What replaced the Bing Search API?
Microsoft retired the Bing Web Search API in August 2025. It points developers to Grounding with Bing Search in Azure AI Foundry. That service gives a model web results to write an answer, and Microsoft's documentation says developers do not receive the raw results. If you need the result list yourself, consider Brave Search API, SerpApi, Exa or Parallel.
What is the difference between a SERP API and an AI search API?
A SERP API returns the results page a search engine would show, as data. An AI search API runs a search, reads the pages and returns passages, sources or an answer for a model to use. Use a SERP API for rank tracking and link lists. Use an AI search API when a model must write from the sources. Many teams use both, which costs more but covers both jobs.
How much does a search API call cost?
Most search APIs charge per 1,000 requests for search alone. Parallel's Fast mode is $1 per 1,000 requests. Brave Search API and Perplexity's standard Search are $5 per 1,000. Exa's search prices run from $4 to $15 per 1,000 depending on type. Credit and token billing can make a full task cost more once page reads and model tokens are added. Compare the cost of one completed task, not one search.
Can a search API ground an AI answer in sources?
Yes, but grounding depends on the whole pipeline. The API must return passages that contain the claim, the model must use them, and your app must show the source link. Test a sample of answers by checking that each cited sentence appears on the cited page. Some providers return quotes for you to check, and others return links only.
Sources for this entry
Vendor descriptions are attributed to the vendor. Prices change; confirm with each vendor.
Continue this reading path
Cite this entry
https://apisearchengine.com/