Skip to content

Search intelligence API
for AI agents

One client for Search, News, Maps, Shopping, Scholar, and more. Structured output your agents can consume without parsing HTML.

Showing Google Search example: ai agents

Google Search

Found results for "ai agents" with structured data, ready for your workflow.

const google = require('@microlink/google')({ apiKey: MICROLINK_API_KEY })

const page = await google('ai agents', { type: 'search' })

console.log(page.results)
[
  {
    "title": "AI agents: The next wave of productivity",
    "url": "https://techcrunch.com/2024/05/30/ai-agents/",
    "source": "TechCrunch",
    "published": "30 May 2024",
    "description": "Leaders discuss the intersection of AI and startup innovation..."
  },
  {
    "title": "How AI agents are reshaping workflows",
    "url": "https://venturebeat.com/ai/ai-agents/",
    "source": "VentureBeat",
    "published": "28 May 2024",
    "description": "Exploring real-world use cases and platform strategies."
  },
  {
    "title": "Building reliable AI agent systems",
    "url": "https://spectrum.ieee.org/ai-agents",
    "source": "IEEE Spectrum",
    "published": "27 May 2024",
    "description": "Best practices for retrieval, context, and evaluation."
  }
]
  • 10 supported search surfaces in one client.
  • Structured results plus LLM-ready Markdown and HTML for top matches.
  • Structured results for prices, ratings, coordinates, and citations.
  • Proxy-backed requests from the first call.
Agentic retrieval

Search first,
fetch later

Search stays lightweight on the first pass so technical workflows can stay fast under real production load.

Ship LLM-ready Markdown

RAG pipelines rarely want raw HTML. They want cleaner text that is easier to embed, rerank, cite, and pass into prompts without wasting context on navigation or markup noise.

Lazy-load the web

Search works best as a two-step system: lightweight results first, deeper content second. That keeps the browse step snappy, then spends the heavier extraction cost only where confidence is already high.

Turn Search into a document discovery engine

Combine operators like site: and filetype: to hunt for papers, docs, filings, changelogs, or PDFs before you enrich anything. That gives technical teams much tighter recall from the first query.

Simple, predictable pricing

One dollar,
one thousand requests

Search has no free tier because reliable result collection depends on managed proxy capacity, regional routing, and production safeguards on every call.

Pro

€39
/month
46,000 requests/month
Managed proxy-backed requests
10 supported search surfaces
Structured normalized results
Location and period controls
Pagination with .next()
Optional page Markdown or HTML enrichment
Automate web discovery

Automate Web Discovery
without scraper debt

Initialize once, choose the surface you need, then paginate or enrich only when a workflow needs more context.

1.Install and initialize

Install @microlink/google, add your Microlink API key, and create one client you can reuse across every supported search surface.

pnpm add @microlink/google

2.Run the first query

Choose the surface you need with the type option and keep the same client shape for search, news, images, maps, shopping, and more.

const google = require('@microlink/google')({
        apiKey: process.env.MICROLINK_API_KEY
      })

      const page = await google('ai agents', {
  type: 'search'
})

console.log(page.results)

3.Lazy-load the web

Keep the first pass fast, then enrich only the winners. Browse lightweight result pages first and call .markdown() or .html() only for the top matches that deserve deeper inspection.

  • Any result with a URL exposes .markdown() for LLM-ready Markdown on demand.
  • Call .html() only when your workflow actually needs raw page markup.
  • Just call .next() to fetch the next page.
  • Lazy-load the web: scan results at ~1s latency, then enrich only the top 3 matches.
Connect everything

Plug Microlink
into your workflow

Combine Search with Metadata, Screenshot, and Markdown to turn discovered URLs into richer outputs for structured fields, visual captures, and AI-ready page content, all under the same paid Microlink plan.

Product Information

Everything you need to know about Microlink Search, pricing, and supported search surfaces.

Microlink Search is a paid search intelligence API for querying and normalizing public results from multiple Google surfaces through one product.

@microlink/google is the Node.js client for integrating Search into your own SEO tooling, monitoring jobs, and AI workflows.

Is this an official Google product?

No. Search is an independent Microlink product that works on top of public Google surfaces.

It is not affiliated with, endorsed by, or provided by Google.

Why is there no free tier?

Search starts on paid plans because reliable public-result collection depends on managed proxy capacity from the first request.

That cost is part of the product itself, so even small workloads use the same proxy-backed delivery model as production workloads.

Which surfaces are supported?

You can query Google Search, Google News, Google Images, Google Videos, Google Places, Google Maps, Google Shopping, Google Scholar, Google Patents, and Google Autocomplete.

Each one keeps the same client shape so teams can ship faster with less parser logic and less provider-specific branching.

What makes this different from a generic SERP API?

Search is designed around normalized output and reusable primitives instead of raw provider-specific payloads.

That means less cleanup for your codebase and faster handoff into rank tracking, market research, or agent pipelines.

How do pagination and HTML enrichment work?

Every result page can call `.next()` to fetch the following page, so pagination can be chained naturally.

Any result containing a URL can also expose `.html()` so you only fetch page markup when a workflow actually needs it.

Is it a fit for SEO and AI workflows?

Yes. Teams use Search for rank tracking, news monitoring, local research, query clustering, citation discovery, and agent enrichment.

The value is consistent structured output plus proxy-backed delivery for recurring public-result collection.

Can I run international or local queries?

Yes. You can use options like `location` and `period` to tune regional intent and recency for multilingual SEO and geo-specific analysis.

That makes the same integration model useful for local search intelligence as well as broader monitoring workflows.

Google is a trademark of Google LLC. Microlink Search is an independent product and is not affiliated with or endorsed by Google.