Exa MCP Server

What is Exa MCP Server?

Exa MCP Server is an implementation of the Model Context Protocol (MCP) that connects LLM-powered agents to Exa AI’s advanced web and data search tools. By running as a lightweight server, it allows applications like Claude Desktop to seamlessly integrate Exa’s search and research APIs, giving AI assistants live access to web search, academic research, company data, crawling, GitHub search, and more—while keeping API keys and data exchange secure and configurable within your infrastructure.

How to Configure Exa MCP Server

Obtain your Exa API key from dashboard.exa.ai/api-keys.
Install the Exa MCP server globally with npm install -g exa-mcp-server or use npx for instant usage.
Open your Claude Desktop configuration file (claude_desktop_config.json).

Add the Exa MCP server configuration under mcpServers, specifying your API key and (optionally) which tools to enable, for example:

{
  "mcpServers": {
    "exa": {
      "command": "npx",
      "args": ["-y", "exa-mcp-server", "--tools=web_search_exa,research_paper_search"],
      "env": {
        "EXA_API_KEY": "your-api-key-here"
      }
    }
  }
}

Save the configuration and fully restart Claude Desktop for changes to take effect.
Alternatively, you can use Exa’s remote MCP server by configuring the MCP server URL instead of running it locally.

How to Use Exa MCP Server

Once configured, launch Claude Desktop and ensure the Exa MCP server shows as connected—look for its icon or status indicator.
When using Claude, simply ask questions or issue tasks that require web search, academic papers, company information, GitHub lookup, or deep research.
The AI assistant will invoke the Exa MCP tools as needed, fetching up-to-date data or performing custom research.
To run the server manually, use npx exa-mcp-server and add --tools to specify available tools. For an overview of tools, run npx exa-mcp-server --list-tools.

Key Features

Real-time integration with Exa AI’s search and research APIs for live data access.
Secure API key management and server operation—keys are never exposed to the AI model.
Flexible tool selection via configuration, with support for specialized searches (academic, company, web, GitHub, deep research, and more).
Simple setup with NPM or remote URL; supports quick start for both technical and non-technical users.
Extensible—quickly add or remove tools as your workflow evolves.

Use Cases

Enabling Claude Desktop or other LLM agents to retrieve the latest web and research information in conversations.
Automating company research, competitor analysis, and LinkedIn lookups directly from your AI workspace.
Running deep research projects where the agent surveys multiple sources, synthesizes findings, and delivers structured reports.
Enriching AI chatbots with GitHub repository data, academic publications, and extracted website content with minimal setup.
Academic or professional environments needing safe, compliant real-time data integration for AI workflows.

FAQ

Q: Why is the Exa MCP server not showing as connected in Claude Desktop?
A: Double-check your configuration in claude_desktop_config.json, ensure your API key is correct, and completely restart Claude Desktop.

Q: What tools are available, and how can I enable only specific ones?
A: Run npx exa-mcp-server --list-tools to see all tools. In your MCP server config, set the --tools argument with a comma-separated list of tool names to enable only those you want.

Q: Where do I get my Exa API key and how should I input it?
A: Get your API key at dashboard.exa.ai/api-keys. Input it in your config file under the EXA_API_KEY environment variable—do not use extra spaces or quotes.

Q: Do I need to restart Claude Desktop after changing the config?
A: Yes, fully quit and relaunch Claude Desktop for any config or tool changes to take effect.

Q: Is it better to run the server locally or use the remote Exa MCP endpoint?
A: Both options are supported—running locally keeps all integration on your device, while the remote endpoint offers instant setup and updates.

Tool Name	Description
web_search_exa	Performs real-time, optimized web searches, retrieving the most relevant search results and extracting content from the web for use in LLM conversations and research assignments.
research_paper_search	Specializes in academic research by searching for scholarly papers, articles, and content from reputable research databases.
company_research	Gathers comprehensive business information by crawling company websites and related sources to create robust company profiles for research and analysis.
crawling	Extracts and retrieves content from specific URLs, such as articles, PDFs, or full web pages, given a precise address.
competitor_finder	Identifies competitors of a business by analyzing industry, offering, and market positioning to surface alternative companies.
linkedin_search	Searches LinkedIn for companies and individuals, enabling targeted people, organization, and profile lookups via name or LinkedIn URLs.
wikipedia_search_exa	Searches and fetches information from Wikipedia, providing quick access to structured entries and knowledge about virtually any topic.
github_search	Enables real-time search across GitHub repositories, issues, and user accounts for development, project research, and code discovery.
deep_researcher_start	Initiates a smart, AI-powered research workflow where the agent investigates complex questions by surveying multiple sources for a thorough, synthesized report.
deep_researcher_check	Checks the status of an ongoing deep research project, returning synthesized findings and a comprehensive research report once complete.

Exa MCP Server is a protocol server that enables AI assistants to securely access real-time search and research tools provided by Exa AI, extending the capabilities of LLM-based applications like Claude Desktop with instant, dynamic web and data retrieval.

What is Exa MCP Server?

How to Configure Exa MCP Server

How to Use Exa MCP Server

Key Features

Use Cases

FAQ