---
title: "Web search APIs"
description: "Compare web search APIs for AI agents and LLMs, including Tavily, Exa, Brave, Perplexity, Parallel, and SerpApi, by index source, output format, and status."
type: "platform-category"
canonical_url: "https://theaiengineering.com/platforms/web-search-apis"
markdown_url: "https://theaiengineering.com/platforms/web-search-apis.md"
topic: "AI agents"
author: "The AI Engineering Editors"
publisher: "The AI Engineering"
citation: "The AI Engineering Editors. \"Web search APIs.\" The AI Engineering. https://theaiengineering.com/platforms/web-search-apis"
---

# Web search APIs

A web search API for AI returns search results in a form a language model can use directly, such as ranked URLs with clean excerpts, full page text, or cited answers, so an agent can ground its output in current information instead of its training data.

Some providers run their own index, including Exa, Brave, Parallel, and Mojeek, while SERP APIs such as SerpApi, Serper, and SearchApi return Google results as structured JSON. Tavily, Linkup, and You.com tune results for agents, and Firecrawl and Jina Reader focus on turning pages into Markdown. SearXNG is an open-source metasearch engine that teams can self-host.

Model providers also offer built-in search tools, including the OpenAI web search tool, the Claude web search tool, and Grounding with Google Search. Microsoft retired the Bing Search APIs in August 2025 in favor of Grounding with Bing Search, and Google's Custom Search JSON API is closed to new customers.

## Web search APIs compared

| Platform | Open source | Deployment | Best for |
| --- | --- | --- | --- |
| [Brave Search API](https://brave.com/search/api/) | No | API | Results from an independent, non-Google index |
| [Bright Data SERP API](https://brightdata.com/products/serp-api) | No | API | Geo-targeted SERP scraping at volume |
| [Claude web search tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool) | No | API | Claude-based agents that need cited web answers |
| [DataForSEO SERP API](https://dataforseo.com/apis/serp-api) | No | API | Bulk SERP collection across search engines |
| [Exa](https://exa.ai) | No | API | Semantic web search over a provider-owned index |
| [Firecrawl Search](https://www.firecrawl.dev) | Yes (AGPL-3.0) | API, Self-hosted | Search plus full-page scraping in one API |
| [Google Custom Search JSON API](https://developers.google.com/custom-search/v1/overview) | No | API | Existing integrations planning a migration |
| [Grounding with Bing Search](https://learn.microsoft.com/en-us/azure/foundry/agents/how-to/tools/bing-tools) | No | API | Azure agents that need web results |
| [Grounding with Google Search](https://ai.google.dev/gemini-api/docs/google-search) | No | API | Gemini apps that need current web facts |
| [Jina Search](https://jina.ai/reader) | Yes (Apache-2.0) | API | Search results delivered as clean Markdown |
| [Kagi Search API](https://help.kagi.com/kagi/api/search.html) | No | API | Kagi users who want its results inside agents |
| [Linkup](https://www.linkup.so) | No | API | Sourced answers with configurable search depth |
| [Mojeek Search API](https://www.mojeek.com/services/search/web-search-api/) | No | API | A non-US independent index for grounding |
| [OpenAI web search tool](https://developers.openai.com/api/docs/guides/tools-web-search) | No | API | OpenAI-based agents that need cited web answers |
| [Parallel](https://parallel.ai) | No | API | Agents that need search plus structured web research |
| [Perplexity Search API](https://docs.perplexity.ai) | No | API | Ranked web results or cited answers from one vendor |
| [SearchApi](https://www.searchapi.io) | No | API | Detailed Google SERP data as JSON |
| [SearXNG](https://docs.searxng.org) | Yes (AGPL-3.0) | Self-hosted | Self-hosted search for local or private agents |
| [SerpApi](https://serpapi.com) | No | API | Structured Google results for agents and SEO tools |
| [Serper](https://serper.dev) | No | API | Low-overhead access to Google results |
| [Tavily](https://www.tavily.com) | No | API | Agent web search with extraction in one API |
| [Valyu](https://www.valyu.ai) | No | API | Agents that search finance and research sources |
| [You.com Web Search API](https://you.com/product/web-search-api) | No | API | Web and news results with full-page content |

## Profiles

### Brave Search API

The Brave Search API serves results from Brave's independent web index, with web, news, image, and video search endpoints and an LLM Context endpoint that returns compact page content intended for language models.

- Company: Brave Software
- Website: https://brave.com/search/api/
- Open source: No
- Deployment: API
- Best for: Results from an independent, non-Google index
- Permalink: https://theaiengineering.com/platforms/web-search-apis#brave-search-api

### Bright Data SERP API

The Bright Data SERP API scrapes search engine results in real time and returns JSON or HTML with geo-targeting, from a company that also runs proxy networks and other web data collection products.

- Company: Bright Data
- Website: https://brightdata.com/products/serp-api
- Open source: No
- Deployment: API
- Best for: Geo-targeted SERP scraping at volume
- Permalink: https://theaiengineering.com/platforms/web-search-apis#bright-data-serp-api

### Claude web search tool

The Claude web search tool gives Claude access to current web content through the Claude API, returning cited sources, with domain allow and block controls and optional dynamic filtering that keeps only relevant results in context.

- Company: Anthropic
- Website: https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool
- Open source: No
- Deployment: API
- Best for: Claude-based agents that need cited web answers
- Permalink: https://theaiengineering.com/platforms/web-search-apis#claude-web-search-tool

### DataForSEO SERP API

The DataForSEO SERP API collects desktop and mobile search results from Google, Bing, and other engines in HTML or JSON, as one part of DataForSEO's suite of SEO data APIs.

- Company: DataForSEO
- Website: https://dataforseo.com/apis/serp-api
- Open source: No
- Deployment: API
- Best for: Bulk SERP collection across search engines
- Permalink: https://theaiengineering.com/platforms/web-search-apis#dataforseo-serp-api

### Exa

Exa is a search API for AI agents built on its own web-scale index and retrieval models, with endpoints for search, page contents, deep search, monitors, and Websets dataset building, plus a hosted MCP server.

- Website: https://exa.ai
- Open source: No
- Deployment: API
- Languages: Python, TypeScript
- Best for: Semantic web search over a provider-owned index
- Permalink: https://theaiengineering.com/platforms/web-search-apis#exa

### Firecrawl Search

Firecrawl's search endpoint queries the web and can return full page content as Markdown or structured data in the same call, alongside its scrape, crawl, and map endpoints, with an AGPL-3.0 version for self-hosting.

- Company: Firecrawl
- Website: https://www.firecrawl.dev
- GitHub: https://github.com/firecrawl/firecrawl
- Open source: Yes (AGPL-3.0)
- Deployment: API, Self-hosted
- Best for: Search plus full-page scraping in one API
- Permalink: https://theaiengineering.com/platforms/web-search-apis#firecrawl-search

### Google Custom Search JSON API

The Custom Search JSON API returns results from a Programmable Search Engine as JSON; it is closed to new customers, existing customers must move off it by January 1, 2027, and Google suggests Vertex AI Search for up to 50 domains.

Closed to new customers, and Google discontinues the API on January 1, 2027.

- Company: Google
- Website: https://developers.google.com/custom-search/v1/overview
- Open source: No
- Deployment: API
- Best for: Existing integrations planning a migration
- Permalink: https://theaiengineering.com/platforms/web-search-apis#google-custom-search-api

### Grounding with Bing Search

Grounding with Bing Search is a tool in Microsoft Foundry Agent Service that gives agents real-time public web results from Bing, and Microsoft named it the replacement for the Bing Search APIs it retired on August 11, 2025.

- Company: Microsoft
- Website: https://learn.microsoft.com/en-us/azure/foundry/agents/how-to/tools/bing-tools
- Open source: No
- Deployment: API
- Best for: Azure agents that need web results
- Permalink: https://theaiengineering.com/platforms/web-search-apis#grounding-with-bing-search

### Grounding with Google Search

Grounding with Google Search is a Gemini API tool that lets Gemini models use real-time Google Search results while generating a response, with citations to the sources used, for applications already built on Gemini.

- Company: Google
- Website: https://ai.google.dev/gemini-api/docs/google-search
- Open source: No
- Deployment: API
- Best for: Gemini apps that need current web facts
- Permalink: https://theaiengineering.com/platforms/web-search-apis#grounding-with-google-search

### Jina Search

Jina's s.jina.ai endpoint runs a web search for a query and returns the top results as LLM-friendly Markdown, using the same open-source Reader that converts any URL to Markdown at r.jina.ai; Jina AI is now part of Elastic.

Elastic acquired Jina AI in October 2025, and the Reader and Search endpoints remain available at jina.ai.

- Company: Elastic
- Website: https://jina.ai/reader
- GitHub: https://github.com/jina-ai/reader
- Open source: Yes (Apache-2.0)
- Deployment: API
- Best for: Search results delivered as clean Markdown
- Permalink: https://theaiengineering.com/platforms/web-search-apis#jina-search

### Kagi Search API

The Kagi Search API gives programmatic access to Kagi's premium search results with an API key and inherits account settings such as blocked or promoted websites, alongside Kagi's FastGPT, Summarizer, and Enrichment APIs.

- Company: Kagi
- Website: https://help.kagi.com/kagi/api/search.html
- Open source: No
- Deployment: API
- Best for: Kagi users who want its results inside agents
- Permalink: https://theaiengineering.com/platforms/web-search-apis#kagi-search-api

### Linkup

Linkup is a web search API for AI with three depth modes, fetch for a single page, search for fast results, and research for in-depth answers, returning sourced answers with full-text snippets and offering private index and bring-your-own-cloud options.

- Website: https://www.linkup.so
- Open source: No
- Deployment: API
- Best for: Sourced answers with configurable search depth
- Permalink: https://theaiengineering.com/platforms/web-search-apis#linkup

### Mojeek Search API

The Mojeek Search API serves results from Mojeek's own independent web index, built by a UK search engine that runs its own crawler, and Mojeek also offers site search, hosted search, and index licensing.

- Company: Mojeek
- Website: https://www.mojeek.com/services/search/web-search-api/
- Headquarters: Brighton, United Kingdom
- Open source: No
- Deployment: API
- Best for: A non-US independent index for grounding
- Permalink: https://theaiengineering.com/platforms/web-search-apis#mojeek-search-api

### OpenAI web search tool

The OpenAI web search tool lets models in the Responses API search the internet for current information before answering and return sourced citations, with filters that allow or block up to 100 domains.

- Company: OpenAI
- Website: https://developers.openai.com/api/docs/guides/tools-web-search
- Open source: No
- Deployment: API
- Best for: OpenAI-based agents that need cited web answers
- Permalink: https://theaiengineering.com/platforms/web-search-apis#openai-web-search-tool

### Parallel

Parallel builds web search and research APIs for AI agents on its own index, with Search for ranked excerpts, Extract for page content, Task for cited deep research, Monitor for change alerts, and FindAll for entity lists.

- Company: Parallel Web Systems
- Website: https://parallel.ai
- Headquarters: Palo Alto, United States
- Founded: 2023
- Open source: No
- Deployment: API
- Best for: Agents that need search plus structured web research
- Permalink: https://theaiengineering.com/platforms/web-search-apis#parallel-web

### Perplexity Search API

The Perplexity Search API returns ranked web results as JSON with titles, URLs, dates, and page text, and Perplexity's Agent API returns cited answers; the Agent API replaced Sonar Chat Completions, whose support ended on September 27, 2026.

- Company: Perplexity
- Website: https://docs.perplexity.ai
- Open source: No
- Deployment: API
- Best for: Ranked web results or cited answers from one vendor
- Permalink: https://theaiengineering.com/platforms/web-search-apis#perplexity-search-api

### SearchApi

SearchApi is a real-time SERP API that scrapes Google organic results, ads, related searches, inline questions, and other result elements and returns them as structured JSON for applications and agents.

- Website: https://www.searchapi.io
- Open source: No
- Deployment: API
- Best for: Detailed Google SERP data as JSON
- Permalink: https://theaiengineering.com/platforms/web-search-apis#searchapi-io

### SearXNG

SearXNG is a free, open-source metasearch engine that aggregates results from many search services without tracking or profiling users, and a self-hosted instance can return JSON results that agents use as a search backend.

- Website: https://docs.searxng.org
- GitHub: https://github.com/searxng/searxng
- Open source: Yes (AGPL-3.0)
- Deployment: Self-hosted
- Languages: Python
- Best for: Self-hosted search for local or private agents
- Permalink: https://theaiengineering.com/platforms/web-search-apis#searxng

### SerpApi

SerpApi is a real-time API that returns Google search results as structured JSON, handling proxies and CAPTCHAs and parsing the rich data on results pages so applications and agents do not need their own scrapers.

- Website: https://serpapi.com
- Open source: No
- Deployment: API
- Best for: Structured Google results for agents and SEO tools
- Permalink: https://theaiengineering.com/platforms/web-search-apis#serpapi

### Serper

Serper is a Google Search API that returns Google search engine results pages as structured JSON, giving applications and LLM agents programmatic access to Google results without running their own scraping infrastructure.

- Website: https://serper.dev
- Open source: No
- Deployment: API
- Best for: Low-overhead access to Google results
- Permalink: https://theaiengineering.com/platforms/web-search-apis#serper

### Tavily

Tavily is a web search API built for AI agents, with endpoints for real-time search, content extraction from URLs, site crawling and mapping, and research tasks, returning results formatted for use in an LLM's context.

Nebius announced its acquisition of Tavily in February 2026, and Tavily now operates as part of Nebius.

- Company: Nebius
- Website: https://www.tavily.com
- Open source: No
- Deployment: API
- Best for: Agent web search with extraction in one API
- Permalink: https://theaiengineering.com/platforms/web-search-apis#tavily

### Valyu

Valyu is a search API that covers the web, financial data, and scientific sources through one endpoint, with a DeepResearch mode that carries out an investigation and returns a cited report.

- Website: https://www.valyu.ai
- Open source: No
- Deployment: API
- Best for: Agents that search finance and research sources
- Permalink: https://theaiengineering.com/platforms/web-search-apis#valyu

### You.com Web Search API

The You.com Web Search API returns real-time web and news results for LLMs and agents, with snippets or full-page content as Markdown or HTML, and an MCP server for clients such as Claude and Cursor.

- Company: You.com
- Website: https://you.com/product/web-search-api
- Open source: No
- Deployment: API
- Best for: Web and news results with full-page content
- Permalink: https://theaiengineering.com/platforms/web-search-apis#you-com-search-api

## How to choose

1. **Index source**: Results come from a provider's own crawl, from scraped Google or Bing results, or from a metasearch of other engines, which affects coverage, freshness, and terms of use.
2. **Output format**: Some APIs return links and snippets, others return full page content as Markdown or a synthesized answer with citations, which changes how many tokens reach the model.
3. **Query controls**: Domain allow and block lists, date and region filters, and search depth settings let an agent narrow results for a task.
4. **Extraction and research**: Crawl, extract, and multi-step research endpoints remove the need for a separate scraper when an agent needs full documents.
5. **Data handling**: Retention terms, zero data retention options, and rules on storing or displaying results matter for regulated or customer-facing products.
6. **Integration**: Official SDKs, an MCP server, and ready-made tools in agent frameworks determine how quickly search can be added to an agent.

## Frequently asked questions

### What is a web search API for AI agents?

A web search API for AI agents is a service that takes a query from an agent and returns current web results formatted for a language model, such as ranked URLs with relevant excerpts, full page text, or a cited answer. The agent calls it as a tool during a task. Tavily, Exa, the Brave Search API, and Parallel are examples.

### Tavily vs Exa: what is the difference?

Tavily is a search API built for AI agents, with endpoints for real-time search, extracting content from URLs, crawling and mapping sites, and research tasks; Nebius announced its acquisition of Tavily in February 2026. Exa runs its own web-scale index and retrieval models, with search, contents, deep search, monitors, and Websets for building datasets. Testing both on your own queries is the reliable way to compare quality.

### What is the replacement for the Bing Search API?

Microsoft retired the Bing Search and Bing Custom Search APIs on August 11, 2025, and named Grounding with Bing Search in Microsoft Foundry Agent Service as the replacement, which works inside Foundry agents rather than as a standalone results API. Teams that need raw results outside Azure usually move to an independent search API, such as the Brave Search API or Exa, or to a SERP API.

### Are there open-source alternatives to Tavily?

Yes. SearXNG is an AGPL-3.0 metasearch engine that aggregates results from many search services, and a self-hosted instance returns JSON when that format is enabled in its settings. Firecrawl publishes an AGPL-3.0 version of its search and scraping API for self-hosting, and Jina Reader is open source under Apache-2.0. Self-hosted options still depend on upstream search engines or on crawling you run yourself.

## Related guides

- [What is tool calling? How LLM function calling works](https://theaiengineering.com/articles/tool-calling): Tool calling lets an LLM request actions from your code through structured calls. Learn the request loop, parallel calls, structured outputs, security, and MCP.
- [What is retrieval-augmented generation (RAG)? How RAG works](https://theaiengineering.com/articles/what-is-rag): Retrieval-augmented generation (RAG) grounds LLM answers in your documents. Learn the RAG pipeline, hybrid search, agentic RAG, evaluation, and failure modes.
- [What is context engineering? A practical guide for AI agents](https://theaiengineering.com/articles/context-engineering): Context engineering is the practice of choosing what fills an LLM context window. Learn its components, why context rot happens, and how to measure results.

Product details change quickly; check each vendor's site for current features, licensing and pricing.

## Cite this page

The AI Engineering Editors. "Web search APIs." The AI Engineering. https://theaiengineering.com/platforms/web-search-apis

- Title: Web search APIs
- Author: The AI Engineering Editors
- Publisher: The AI Engineering
- URL: https://theaiengineering.com/platforms/web-search-apis
- Markdown: https://theaiengineering.com/platforms/web-search-apis.md
