Built by Metorial, the integration platform for agentic AI.

Learn More

exa

Exa

    Server Summary

    • Neural web search

    • Semantic content discovery

    • Similarity-based research

    • AI content summarization

Exa MCP Server

The Exa MCP Server provides seamless integration with Exa's powerful neural search engine, enabling intelligent web searching, content discovery, and similarity-based research directly within your MCP-compatible applications. This server harnesses Exa's advanced AI capabilities to deliver high-quality search results with semantic understanding, making it ideal for research, content analysis, and information retrieval tasks. Whether you're conducting deep research, finding related articles, or extracting content from specific URLs, this server provides the tools you need to access and analyze web content intelligently.

Overview

Exa is a next-generation search engine that uses neural networks to understand the semantic meaning of queries rather than just matching keywords. This MCP server exposes Exa's core capabilities through three powerful tools and corresponding resource templates, allowing you to search the web intelligently, discover similar content, and retrieve full page content with AI-generated summaries and highlights.

Unlike traditional keyword-based search engines, Exa's neural search understands context and intent, making it particularly effective for research, competitive analysis, content discovery, and finding authoritative sources on complex topics. The server supports extensive filtering options including date ranges, domain restrictions, content patterns, and categories, giving you precise control over your search results.

Features

Neural Search Intelligence

Exa's neural search engine goes beyond simple keyword matching to understand the semantic meaning and context of your queries. This enables more relevant results, especially for complex or conceptual searches where traditional search engines might struggle. The search engine can automatically optimize queries using autoprompt, and supports both neural and keyword search modes.

Content Extraction and Summarization

Retrieve full text content from web pages along with AI-generated summaries and relevant highlights. This makes it easy to quickly understand the key information from multiple sources without manually reading through entire articles. The server can process up to 100 URLs simultaneously, making bulk content analysis efficient and straightforward.

Similarity Discovery

Find related content based on any URL, enabling powerful discovery workflows. This is particularly useful for literature reviews, competitive analysis, finding related research papers, or discovering content on similar topics. The similarity search uses neural embeddings to find genuinely related content rather than just pages with similar keywords.

Flexible Filtering

Comprehensive filtering options allow you to narrow results by publication date, include or exclude specific domains, filter by content patterns, and categorize results. This precision ensures you get exactly the information you need without wading through irrelevant results.

Tools

exa_search

Search the web using Exa's neural search engine with support for advanced filtering and content options.

Parameters:

  • query (required, string): The search query to execute
  • searchType (string): Choose between "neural" for semantic search, "keyword" for exact matching, or "auto" to let Exa decide. Default is "auto"
  • numResults (number): Number of results to return, between 1 and 100. Default is 10
  • useAutoprompt (boolean): Enable automatic query optimization to improve search relevance
  • category (string): Filter results by category such as "research paper", "news", or "company"
  • includeDomains (array of strings): Restrict results to only these domains
  • excludeDomains (array of strings): Exclude results from these domains
  • startPublishedDate (string): Only include results published after this date in ISO 8601 format
  • endPublishedDate (string): Only include results published before this date in ISO 8601 format
  • includeTextPatterns (array of strings): Only include results containing these text patterns
  • excludeTextPatterns (array of strings): Exclude results containing these text patterns
  • includeText (boolean): Include the full text content from pages. Default is true
  • includeSummary (boolean): Include AI-generated summaries of the content
  • includeHighlights (boolean): Include relevant highlights extracted from the content

exa_find_similar

Discover web pages similar to a given URL using neural similarity search.

Parameters:

  • url (required, string): The URL to find similar content for
  • numResults (number): Number of similar results to return, between 1 and 100. Default is 10
  • category (string): Filter similar results by category
  • excludeSourceDomain (boolean): Exclude results from the same domain as the source URL
  • startPublishedDate (string): Only include results published after this date in ISO 8601 format
  • endPublishedDate (string): Only include results published before this date in ISO 8601 format
  • includeText (boolean): Include the full text content from pages. Default is true
  • includeSummary (boolean): Include AI-generated summaries of the content
  • includeHighlights (boolean): Include relevant highlights extracted from the content

exa_get_contents

Retrieve full content, metadata, and AI-generated summaries for specific URLs in bulk.

Parameters:

  • urls (required, array of strings): Array of URLs to retrieve content for. Must include between 1 and 100 URLs
  • includeText (boolean): Include the full text content from pages. Default is true
  • includeSummary (boolean): Include AI-generated summaries of the content
  • includeHighlights (boolean): Include relevant highlights extracted from the content
  • subpages (number): Number of subpages to include for each URL, between 0 and 10
  • livecrawl (string): Control live crawling behavior. Options are "always" to force fresh crawling, "fallback" to use live crawling if cached content is unavailable, or "never" to only use cached content

Resource Templates

exa-search

Access search results for a specific query through the resource system.

URI Template: exa://search/{query}

The query parameter should be URL-encoded. This resource template provides a convenient way to access and cache search results for repeated queries.

exa-content

Access the full content and metadata for a specific URL through the resource system.

URI Template: exa://content/{url}

The URL parameter should be URL-encoded. This template allows you to retrieve and reference content from specific URLs consistently.

exa-similar

Access similar content results for a given URL through the resource system.

URI Template: exa://similar/{url}

The URL parameter should be URL-encoded. Use this template to find and reference similar content based on any URL.

Use Cases

Research and Literature Review

Use the neural search capabilities to find relevant research papers, articles, and authoritative sources on complex topics. The similarity search is particularly powerful for discovering related papers and building comprehensive literature reviews.

Content Discovery and Curation

Discover high-quality content on specific topics using semantic search. The AI-generated summaries and highlights make it easy to quickly evaluate multiple sources and identify the most relevant content for your needs.

Competitive Analysis

Find similar websites, articles, or companies using the similarity search. Combine with domain filtering to analyze specific industry segments or exclude certain competitors.

Information Extraction

Retrieve and process content from multiple URLs simultaneously with full text extraction, summaries, and highlights. This is ideal for building knowledge bases, monitoring specific sources, or analyzing content trends.

Time-Sensitive Research

Use the date filtering capabilities to focus on recent content or historical information within specific time periods. This is particularly useful for news monitoring, trend analysis, or tracking the evolution of topics over time.

Best Practices

When using neural search, phrase your queries conceptually rather than as simple keywords. For example, instead of "machine learning papers", try "recent advances in deep learning architectures for computer vision". The neural search engine understands context and will return more relevant results for well-articulated queries.

Enable autoprompt when you're not sure how to phrase your query optimally. This feature automatically reformulates your query to improve search relevance.

Use the similarity search to explore topics deeply. Start with a high-quality article or paper on your topic, then use find_similar to discover related content. This often yields better results than trying to formulate the perfect search query.

Take advantage of the content extraction features by enabling summaries and highlights. This allows you to quickly process multiple sources and identify the most valuable content without reading every page in full.

Combine multiple filtering options to narrow results precisely. For example, you might search for content within specific domains, published within a date range, and containing certain text patterns to find exactly what you need.