# Serper MCP

> Serper is a dedicated MCP Server that connects your AI agent to real-time Google search data. It lets you run three distinct types of searches programmatically: general organic search results (`google_search`), the latest news articles (`google_news_search`), and visual image queries (`google_image_search`). You get structured, actionable data—not just a web page dump—perfect for research pipelines or content aggregation.

## Overview
- **Category:** developer-tools
- **Price:** Free
- **Tags:** search-api, real-time-data, news-aggregation, image-search

## Description

You need an API hook into real Google Search data for your agent? Forget scraping HTML; this **Serper** MCP Server gives your AI client direct, structured access to what's happening on the web. It lets you run three distinct kinds of searches programmatically: general organic results using `google_search`, tracking breaking stories with `google_news_search`, or pulling visual assets via `google_image_search`. You get clean data—titles, links, snippets—not just a wall of text that breaks your pipeline. 

When you run a standard search query through the `google_search` tool, you're not getting vague results; you're instantly receiving structured SERP data for organic web hits. This means your agent gets titles, direct links, descriptive snippets, and even the ranking position of those results. You can specify where you want to search from by using geolocation or language parameters, so the output always matches the context you need.

If you're tracking news—the kind that breaks overnight—you use `google_news_search`. This tool searches across Google News specifically and hands back a list of recent articles. For each hit, you get the headline, the source publication, the exact date it was published, and a summary of what it's about. It keeps your agent updated on breaking stories without having to monitor dozens of different RSS feeds.

For any visual research pipeline—maybe you need reference photos or asset sourcing for content—you use `google_image_search`. This tool runs an image query and returns structured data, listing the specific image URLs, titles, and where those images originated from. It's perfect when your agent needs to verify visuals or gather source material that requires a concrete link.

It’s all about actionable data streams. You run `google_search` for standard results, getting links, snippets, and positions instantly; you use `google_news_search` to pull recent articles with their sources, dates, and summaries; and you tap into `google_image_search` when you need structured URLs, titles, and source pages. Your agent doesn't just 'check' something—it gets precise data points it can actually work with for research or content aggregation. You connect this server, input your key, and you’ve got three powerful search tools running right off the bat.

## Tools

### google_image_search
Queries Google Images and returns structured results, giving you image URLs, titles, and source pages for visual research.

### google_news_search
Searches Google News to return a list of recent articles matching your query, including headlines, sources, dates, and summaries.

### google_search
Executes a real-time Google Search for organic SERP results, providing titles, links, snippets, and positions instantly.

## Prompt Examples

**Prompt:** 
```
Search Google for 'best AI agent frameworks 2026' and show me the top 5 results.
```

**Response:** 
```
Found 5 results for 'best AI agent frameworks 2026':

| # | Title | Source |
|---|---|---|
| 1 | Top 10 AI Agent Frameworks | datacamp.com |
| 2 | LangGraph vs CrewAI vs AutoGen | towardsdatascience.com |
| 3 | Best Agent Tools 2026 | dev.to |
| 4 | Building AI Agents Guide | langchain.com |
| 5 | Agent Framework Comparison | medium.com |
```

**Prompt:** 
```
Search the latest news about OpenAI.
```

**Response:** 
```
Found 10 news articles about OpenAI:

| # | Title | Source | Date |
|---|---|---|---|
| 1 | OpenAI Launches New Agent API | techcrunch.com | 2h ago |
| 2 | GPT-5 Release Timeline | theverge.com | 5h ago |
| 3 | OpenAI Valued at $300B | bloomberg.com | 1d ago |
```

**Prompt:** 
```
Search Google Images for 'neural network architecture diagram'.
```

**Response:** 
```
Found 10 image results for 'neural network architecture diagram':

1. **CNN Architecture Diagram** — researchgate.net
2. **Transformer Model Architecture** — arxiv.org
3. **Deep Learning Layers Visualization** — towardsdatascience.com
```

## Capabilities

### Perform real-time general web searches
You query Google for standard organic results, receiving structured titles, links, snippets, and ranking positions.

### Track breaking news headlines
The agent executes a search across Google News, pulling recent articles with their source, date published, and summary.

### Gather visual research assets
You perform an image search and get structured results listing the image URLs, titles, and where they originated.

## Use Cases

### Tracking competitor launches
A product manager needs to know if a rival just released a new feature or updated their site. They ask their agent: 'Check the latest news and general search for Acme Corp.' The agent calls `google_news_search` first, then runs `google_search`. This confirms any recent PR announcements or major keyword shifts.

### Building a content brief
A technical writer is starting an article on 'quantum computing.' They tell their agent: 'Find key concepts and visuals.' The agent runs `google_search` to get top articles, then uses `google_image_search` to gather diagrams (e.g., circuit diagrams) for illustration ideas.

### Investigating local market trends
An international sales team needs to understand consumer sentiment in Brazil regarding a new product category. They tell their agent: 'Search Google for X, but set the language and geo parameters for Brazil.' The agent uses `google_search` with localized settings to pull relevant search snippets.

### Academic literature review
A data scientist needs visual examples of specific network architectures. They prompt their agent: 'Show me diagrams related to transformer models.' The agent executes `google_image_search`, gathering URLs and sources from reliable academic sites for the final paper.

## Benefits

- **Real-time Data:** You get instant access to live Google results. Stop relying on cached data; use `google_search` for up-to-the-minute insights.
- **Structured Output:** The system doesn't return HTML garbage. It delivers clean JSON objects, so your agent can immediately parse titles, links, and snippets using any tool.
- **Multifaceted Research:** Need both the news and the visuals? Your agent runs `google_news_search` for context and then uses `google_image_search` to gather supporting diagrams in one flow.
- **Localization Control:** Don't just search globally. Use parameters with `google_search` to target specific geographic locations (gl) or languages (hl).
- **Cost-Effective Scaling:** The free tier provides 2,500 searches monthly. This keeps initial testing cheap while giving you enough volume for serious small-scale projects.

## How It Works

The bottom line is: you tell your AI client what to search for, and it gets back organized, machine-readable data—not a messy webpage.

1. Subscribe to the Serper server and provide your API key within your AI client.
2. Your agent calls a tool (e.g., `google_search`) with specific parameters like query, location, or language code.
3. Serper processes the request against Google's APIs and returns clean, structured JSON data directly to your agent.

## Frequently Asked Questions

**How many free searches do I get?**
Serper offers 2,500 free searches per month — no credit card required. This includes Google Search, News, and Images queries. Paid plans start at $50/month for 50,000 searches with additional volume discounts.

**Can I get results for specific countries and languages?**
Yes! Serper supports geolocation (gl parameter) and language (hl parameter) for all search types. For example, set gl='br' and hl='pt' for Brazilian Portuguese results, or gl='fr' and hl='fr' for French results. This works for Search, News, and Images endpoints.

**What types of search results are available?**
Serper provides three search types: organic web results (with titles, links, snippets, and knowledge panels), Google News (with headlines, sources, dates), and Google Images (with image URLs and source pages). All results include structured metadata for easy parsing.

**How do I set up Serper in my AI agent?**
You must subscribe to the server and provide your unique API key. Once configured, your AI client connects instantly and gains access to all three search tools.

**What format does `google_search` deliver its data in?**
It returns structured JSON objects. You get titles, links, snippets, and positions—clean data your agent processes reliably without needing to parse messy HTML.

**How fast are the results when using `google_news_search`?**
The latency is sub-100ms. This speed ensures your AI client receives real-time news data instantly, supporting quick decision-making workflows.

**What if my usage hits a rate limit with Serper?**
You will receive standard HTTP error codes and clear messages detailing the issue. The system pauses your requests until the allotted time passes, protecting against data loss.

**Can `google_image_search` work across different AI frameworks?**
Yes, it works with major frameworks like LangChain and CrewAI. Serper's MCP standard guarantees seamless integration regardless of your agent’s core language.