# Exa MCP MCP

> Exa. This MCP gives your AI client semantic web search—it finds concepts and meaning, not just keywords. Instead of relying on exact matches that miss context, Exa understands natural language queries across the entire web. You can use it to find related articles for research or quickly pull clean text from specific URLs.

## Overview
- **Category:** ai-frontier
- **Price:** Free
- **Tags:** semantic-search, neural-search, web-indexing, information-retrieval, ai-data-sourcing, contextual-search

## Description

Connect this MCP to your AI client when simple keyword searching isn't cutting it. Standard search engines only match what you type; they don't understand the underlying idea. Exa changes that by performing semantic searches, finding results based on concepts and meaning. Need to build a knowledge base from several sources? You can feed it specific URLs, and it pulls clean text, key highlights, and summaries for every page. Want to know what your competitors are talking about? Give it a single link, and Exa finds pages that cover the same topic or subject matter. Because Vinkius hosts this MCP in its catalog, you connect once—and get access to deep web understanding whether you're working inside an IDE or running an agent script.

## Tools

### exa_find_similar
Finds web pages that cover the same topic or are conceptually related to a specific URL.

### exa_get_contents
Extracts clean text, key highlights, and summaries from multiple specified URLs.

### exa_search
Searches the web using conceptual understanding to find results relevant to a given query.

## Prompt Examples

**Prompt:** 
```
Search for companies building memory infrastructure for AI agents.
```

**Response:** 
```
Found 10 semantically relevant results:

1. **Mem0** (mem0.ai) — Score: 0.94
   Leading memory layer for AI agents with hybrid KV+vector+graph storage...

2. **Zep** (getzep.com) — Score: 0.91
   Long-term conversational memory with temporal knowledge graphs...

3. **Letta** (letta.com) — Score: 0.88
   OS-inspired agent runtime with self-managed memory hierarchy...
```

**Prompt:** 
```
Find pages similar to https://docs.langchain.com/docs/get_started/introduction
```

**Response:** 
```
Found 10 pages similar to LangChain's introduction:

1. **LlamaIndex Documentation** (docs.llamaindex.ai) — Score: 0.92
2. **CrewAI Getting Started** (docs.crewai.com) — Score: 0.89
3. **AutoGen Documentation** (microsoft.github.io/autogen) — Score: 0.86
4. **Haystack Introduction** (docs.haystack.deepset.ai) — Score: 0.83
```

**Prompt:** 
```
Extract the content from these 3 URLs: https://arxiv.org/abs/2401.00001, https://openai.com/blog, https://anthropic.com/research
```

**Response:** 
```
Extracted content from 3 URLs:

### 1. arxiv.org — Research Paper
> Key highlights: Novel approach to multi-agent coordination using graph neural networks...

### 2. openai.com — Blog Post
> Full text extracted: 4,200 words covering latest research developments...

### 3. anthropic.com — Research Page
> Highlights: Constitutional AI methodology and safety-first approach...
```

## Capabilities

### Scope Web Concepts
Search the entire web using natural language queries that find conceptually relevant results.

### Map Related Content
Identify and return web pages with similar topics or themes based on a single source URL.

### Extract Structured Data
Pull clean text, key highlights, and summaries from a batch of specific URLs into an organized format.

## Use Cases

### Mapping Out Topic Coverage
A content team wants to write an article on 'quantum computing for medicine.' They use `exa_search` first to get 20 core articles. Then, they run those links through `exa_find_similar` to locate adjacent topics—like 'bio-inspired algorithms' or 'drug delivery systems'—and build out a full outline.

### Building an Internal Research Database
A developer needs to ingest the latest findings from three major industry blogs. They use `exa_get_contents` by listing the URLs, which pulls structured text and highlights, ready for immediate vector indexing.

### Competitive Deep Dive
You suspect a competitor is using similar technology to yours. You feed their product documentation URL into `exa_find_similar`, and the MCP returns several other technical white papers that discuss the same underlying concepts, giving you an edge.

### Validating Assumptions
You're writing a paper on AI ethics. You use `exa_search` to check if 'algorithmic bias in healthcare' is still a hot topic. The results give immediate scorecards and snippets from the latest sources, confirming your angle before you write a single word.

## Benefits

- Don't waste time on standard searches. `exa_search` understands context, giving your agent results based on the *idea* you're researching, not just the words.
- When building a knowledge base, use `exa_get_contents`. Instead of dealing with messy HTML, you get clean text and highlights from multiple sources in one go.
- Need competitive insight? Pass a URL to `exa_find_similar` and immediately discover alternative articles or research papers on the same subject. It's perfect for mapping out your content strategy.
- It drastically cuts down on manual web exploration. You stop opening dozens of tabs just to verify a concept and let the MCP do the heavy lifting.
- The different search types in `exa_search` (auto, fast, deep) give you control over depth versus speed when scoping out a topic.

## How It Works

The bottom line is that you get web context and clean, usable text without writing complex scraping scripts.

1. First, subscribe to this MCP in Vinkius and provide your API key.
2. Your agent sends the query or list of links to the Exa tools via the MCP connection.
3. Exa processes the request—whether it's a broad search, finding similar pages, or extracting content—and returns structured data directly to your AI client.

## Frequently Asked Questions

**How does exa_search differ from a standard Google search?**
It understands concepts, not keywords. Standard searches only match text you type; Exa finds results based on the meaning behind your query, giving you far deeper context.

**What if I want to pull data from 20 different articles?**
Use `exa_get_contents`. You provide a list of URLs, and it pulls clean text, highlights, and summaries for all of them in one operation.

**Can exa_find_similar help me with competitive analysis?**
Yes. Give it a competitor's URL; `exa_find_similar` will return articles that discuss the same subject, helping you map out their full content territory.

**Does this MCP require me to be an AI agent developer?**
No. You just need your preferred AI client connected via Vinkius. It works whether you're writing code or just prompting a conversationally capable agent.

**What happens if I exceed the monthly usage limit when calling exa_search?**
Your agent will receive a rate limit error. The MCP documentation explains how to request an API key upgrade or manage your current quota through Vinkius.

**How do I properly authenticate my agent when using exa_get_contents?**
You must provide your unique Exa API key. The Vinkius platform handles secure credential storage, so just pass the required key as specified by the tool definition.

**Does exa_search provide structured data beyond simple text snippets?**
Yes, it returns more than plain text. Each result includes a relevance score and specific generated highlights, helping your agent rank information accurately.

**Which performance mode should I use when executing exa_search?**
It depends on whether you need speed or depth. Use 'instant' for fast checks, or use the 'deep' setting if you require the most thorough results possible.