WebSearch
APIs
Submit an API

41 APIs · updated Oct 2026

Every search API on earth.

Bing is dead. Google rations theirs. A wave of startups wants to feed your agent the web. Here's who they all are and what they actually return.

Live
40
Own index
13
Full content
11
Answers
18

TL;DR

No time? These are our picks.

Fastest Exa

Exa Fast reports a median latency under 500 ms.

Most accurate Keirolabs

96.5% on SimpleQA in deep mode, 93% in lite.

Value for money KeirolabsParallel

Keirolabs: $0.10 per 1,000 lite calls with 93% reported on SimpleQA. Parallel: $1 per 1,000 on its fast tier.

Most token-efficient Parallel

Returns ranked, compressed excerpts instead of whole pages, so your model reads less.

Speed and accuracy figures are reported by the vendors themselves. See every SimpleQA claim and whether it can be checked.

Which one should you buy?

Three questions. Your shortlist updates as you answer.

1Who reads the results?
2What do you need back?
3What matters most?

Your shortlist

  1. Answer the three questions and your top five shows up here.

Scored from the directory's data. Treat it as a starting point, then run your own bake-off.

016 APIs

Independent index

They crawl the web themselves. No Google or Bing underneath, so nobody upstream can pull the plug.

0212 APIs

AI-native

Built for agents. They fetch the pages, strip the junk and hand your model clean text, sometimes an answer.

0313 APIs

SERP scraper

Google's results page, parsed into JSON. Cheap and familiar, and the scraping is the vendor's problem.

0410 APIs

Big tech

Search as a built-in tool for one company's models. One line to turn on, hard to take with you.

Same question, four APIs

One query.
Four different answers.

One gives you links. One gives you the pages. One writes the answer. One hands you Google's results page as JSON. Picking a search API is mostly picking which of these you want.

curl -G https://api.search.brave.com/res/v1/web/search \
  --data-urlencode "q=latest rust release" \
  -H "Accept: application/json" \
  -H "X-Subscription-Token: $BRAVE_KEY"
Returns → Links + snippets

The directory

41 entries. Click a row for details and alternatives.

APICategoryOwn indexPage contentAnswersFree tier
Brave Search API One of the few Western independent indexes at scale. Has plans aimed at AI grounding. Independent index Yes
Mojeek UK crawler with its own index and a privacy focus. Smaller index, but it is truly independent. Independent index Trial
Kagi Search API Invite-only beta. Blends its own index with third-party sources. Independent index No
Yandex Search API Russia's largest engine, served through Yandex Cloud. Text, image and generative search. Terms differ for non-residents. Independent index Trial
Naver Search API Korea's dominant engine. Separate endpoints for web, news, blogs and more. Best for Korean-language queries. Independent index Daily quota
Marginalia Search Open-source, one-person crawler focused on the small, non-commercial, text-heavy web. Not a general-purpose index. Independent index Shared key
Exa Neural (embedding-based) search over its own index. Good for 'find pages like this' queries. AI-native Credits
Keirolabs Its own index, plus a Google SERP endpoint. Search, extract, answer and research endpoints. Ships an MCP server. AI-native 750 credits
Tavily Built for LLM agents. Returns cleaned content and an optional answer. AI-native Monthly credits
Linkup Paris-based. Offers standard and deep search modes, plus sourced answers. AI-native Credits
Parallel Search Search and web research APIs designed for agents, with ranked, token-dense excerpts. AI-native Credits
Valyu Combines web search with proprietary and academic datasets. AI-native Credits
Perplexity Search API Raw ranked results, plus the Sonar models that return grounded answers. AI-native No
You.com API Web and news search, plus research and answer endpoints. AI-native Trial
Jina Search Add the s.jina.ai prefix to a query and get the top results back as LLM-ready markdown. AI-native Tokens
Firecrawl Search Search, then scrape each result in the same call. Mostly a crawler company. AI-native Credits
TinyFish Search and fetch for agents, currently free under an hourly cap. Also sells browser automation. AI-native Free (rate-limited)
Bocha Chinese web search API that is popular with domestic LLM apps. Web search plus an AI search mode with structured cards. AI-native No
SerpApi The veteran here. Parses Google, Bing, Baidu, YouTube, Maps and many more into JSON. SERP scraper Small plan
Serper Fast, cheap Google SERP JSON. A popular default in LangChain-era stacks. SERP scraper Signup credits
SearchApi Real-time SERP scraping across many engines and verticals. SERP scraper Trial
DataForSEO SERP Pay-as-you-go. Its queued mode is very cheap if you can wait for results. SERP scraper Trial
Bright Data SERP API Enterprise proxy network behind it. Geo-targeting down to city level. SERP scraper Trial
Scrapingdog Budget SERP and scraping API. SERP scraper Trial
Oxylabs SERP Scraper API Enterprise proxy giant. Parsed results from Google, Bing, Baidu and Yandex. SERP scraper Trial
Decodo SERP API Formerly Smartproxy. Comes close to premium providers on speed for a lot less money. SERP scraper Trial
ScraperAPI A general scraping API with structured Google endpoints. Priced in credit bundles, not per search. SERP scraper Trial
Zenserp A simple Google, Bing and YouTube SERP API that has been around for years. SERP scraper Small plan
SerpWow Traject Data's multi-engine SERP API, with batch scheduling for large jobs. SERP scraper Trial
HasData Google SERP API Google SERP plus Maps, News and Shopping endpoints, from a scraping-tools company. SERP scraper Trial
Apify Google Search Scraper A ready-made Actor on the Apify platform. Pay per result, and you can chain it with crawlers in the same workflow. SERP scraper Monthly credits
Google Programmable Searchlimited Closed to new customers. Full-web search shuts down on January 1, 2027, and new engines are capped at 50 domains. Big tech Daily quota
Gemini Grounding Google Search as a tool inside Gemini calls. You get answers and citations, not raw results. Big tech Quota
OpenAI Web Search A tool in the Responses API. Billed per call plus tokens. Only works through OpenAI models. Big tech No
Claude Web Search A server-side tool in the Messages API, with citations. Billed per search plus tokens. Big tech No
Vertex AI Searchlimited Google's suggested replacement for Custom Search. Limited to up to 50 domains, so it is not public whole-web results. Big tech Trial
Grounding with Bing What replaced the Bing API. Only works as a tool for agents inside Azure AI Foundry. Big tech No
Amazon Nova Web Grounding A built-in system tool for Nova models on Bedrock, with citations. Available in US regions only. Big tech No
Mistral Web Search A built-in web_search tool in Mistral's Agents and Conversations APIs. Still in beta. Big tech No
xAI Web Search Grok answers grounded in the web, with citations. The same key can also search posts on X. Big tech No
Bing Web Search APIretired Retired August 11, 2025. Microsoft's replacement is grounding inside Azure AI Agents. Big tech —

No prices on purpose: they change monthly, and every detail page links to the vendor. Wrong about something? Send a correction.