Exa AI Search

Enable AI agents to perform intelligent web searches using Exa’s neural search engine with semantic understanding

The Exa MCP server provides AI agents with access to Exa’s neural search engine, which understands the meaning behind queries rather than just matching keywords. This enables more accurate and relevant search results for complex research tasks.

Overview

Exa’s search capabilities differ from traditional search engines:

• Neural search: Understands query semantics, not just keywords
• Content extraction: Returns clean, structured content from pages
• Company research: Deep crawling of company websites
• LinkedIn integration: Search LinkedIn profiles and companies
• Deep research: Multi-step research with comprehensive reports

Deploy to Heroku

Deploy the Exa MCP server to Heroku with one click:

After deployment, set your Exa API key:

heroku config:set EXA_API_KEY=your-exa-api-key -a your-app-name

Get your API key from dashboard.exa.ai/api-keys.

Register with Heroku AI

Register the MCP server with your Heroku AI model:

heroku ai:mcp:servers:add exa-search \
  --app your-inference-app \
  --server-app your-exa-mcp-app

Available Tools

Web Search

Tool	Description
web_search_exa	Perform semantic web searches with content extraction
crawling	Extract content from specific URLs

Research Tools

Tool	Description
company_research	Comprehensive company research by crawling company websites
linkedin_search	Search LinkedIn for companies and people
deep_researcher_start	Start an AI-powered deep research task
deep_researcher_check	Check status and get results of deep research

Using with Heroku AI Agents

Perform intelligent searches through the Heroku Agents API:

Python

import os
from openai import OpenAI

client = OpenAI(
    base_url=os.getenv("INFERENCE_URL") + "/v1",
    api_key=os.getenv("INFERENCE_KEY")
)

response = client.chat.completions.create(
    model=os.getenv("INFERENCE_MODEL_ID"),
    messages=[
        {"role": "user", "content": "Find recent research papers about transformer architectures in NLP"}
    ],
    extra_body={
        "heroku": {
            "mcp_servers": ["exa-search"]
        }
    }
)

print(response.choices[0].message.content)

cURL

curl $INFERENCE_URL/v1/agents/heroku \
  -H "Authorization: Bearer $INFERENCE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "'$INFERENCE_MODEL_ID'",
    "messages": [
      {"role": "user", "content": "Find recent research papers about transformer architectures in NLP"}
    ],
    "mcp_servers": ["exa-search"]
  }'

Tool Details

web_search_exa

Performs semantic web searches optimized for AI consumption. Returns relevant results with extracted content.

Parameters:

• query (string): The search query
• num_results (integer, optional): Number of results to return (default: 10)
• include_domains (array, optional): Limit search to specific domains
• exclude_domains (array, optional): Exclude specific domains from results

Example query:

"Find tutorials on building RAG applications with LangChain"

company_research

Crawls company websites to gather comprehensive information about businesses.

Parameters:

• company_name (string): Name of the company to research
• company_url (string, optional): Company website URL

Example query:

"Research Anthropic and tell me about their products and team"

crawling

Extracts clean content from specific URLs. Useful for reading articles, documentation, or any web page.

Parameters:

• url (string): The URL to crawl and extract content from

Example query:

"Extract the content from this blog post: https://example.com/blog/ai-trends"

linkedin_search

Searches LinkedIn for people and companies using Exa’s neural search.

Parameters:

• query (string): Search query for LinkedIn profiles or companies

Example query:

"Find the LinkedIn profile of the CEO of Stripe"

deep_researcher_start / deep_researcher_check

For complex research questions, the deep researcher performs multi-step research:

1. Searches multiple sources
2. Reads and analyzes content
3. Synthesizes a comprehensive report

Start a research task:

"Start deep research on the current state of quantum computing and its potential applications in cryptography"

Check research status:

"Check if my quantum computing research is complete"

Configuration Options

Configure which tools are available via environment variables:

heroku config:set ENABLED_TOOLS=web_search_exa,company_research,crawling -a your-exa-mcp-app

Available tools:

• web_search_exa
• company_research
• crawling
• linkedin_search
• deep_researcher_start
• deep_researcher_check

If not specified, all tools are enabled by default.

Example Use Cases

Academic Research

"Find peer-reviewed papers on climate change mitigation strategies published in the last 2 years"

Market Research

"Research the top 3 competitors in the project management software space and summarize their key features"

Technical Documentation

"Find the official documentation for implementing OAuth 2.0 with Python"

News and Trends

"Search for recent news about AI regulation in the European Union"

Company Intelligence

"Research Heroku and provide an overview of their platform, pricing, and target customers"

Neural Search vs Keyword Search

Exa’s neural search understands query intent:

Query	Keyword Search	Neural Search
”How do transformers work”	Results about electrical transformers	Results about transformer neural networks
”Apple latest news”	Mixed results	Results about Apple Inc. technology news
”Python best practices”	General Python results	Coding style guides and tutorials

This semantic understanding makes Exa particularly effective for research and technical queries.

Additional Resources

• Exa Documentation
• Exa API Reference
• Custom Tools Guide
• Heroku Agents API