Exa MCP
Exa MCP Server connects AI assistants to Exa AI's powerful real-time web and code search tools, enabling accurate and efficient coding and research. It supports multiple tools like live web search, code context retrieval, and deep research for developers and researchers.
How to Install and Use Exa MCP
If you want a fast and efficient way to add smart web and code search to your AI assistants like Claude or development tools, then Exa MCP is the right choice. It helps your coding agents find perfect and live web content or code snippets from billions of sources. Here’s how to install and get started with Exa MCP in a few easy steps.
Step 1: Install Exa MCP with NPM or NPX
First, you need to install the Exa MCP server on your computer. The easiest way is to use the Node.js package manager, npm. If you just want to run it directly without installing permanently, you can use npx instead.
Here are the commands to run in your terminal or command prompt:
- To install globally using npm:
npm install -g exa-mcp-server
- To run using npx without installing:
npx exa-mcp-server
By default, this will start Exa MCP with two main tools enabled: web_search_exa and get_code_context_exa.
Step 2: Configure Exa MCP for Your AI or Development Tool
Once installed or running, you need to tell your AI assistant (like Claude, Cursor, or Codex) to use Exa MCP. You do this by adding a simple configuration.
Here’s an example configuration for Claude Desktop:
{
"mcpServers": {
"exa": {
"command": "npx",
"args": ["-y", "exa-mcp-server"],
"env": {
"EXA_API_KEY": "your-api-key-here"
}
}
}
}
Replace "your-api-key-here" with your actual API key from the Exa dashboard at dashboard.exa.ai/api-keys.
If you use Cursor or Claude Code, the setup uses an HTTP configuration like this:
{
"mcpServers": {
"exa": {
"type": "http",
"url": "https://mcp.exa.ai/mcp",
"headers": {}
}
}
}
Step 3: Choose Which Tools to Enable
Exa MCP supports many helpful tools to improve web search, coding help, research, and more. By default, only web_search_exa and get_code_context_exa are turned on. You can enable more as needed.
Here’s how you add tools when running the server with npx:
- Run only the code search tool, recommended for developers:
npx exa-mcp-server tools=get_code_context_exa
- Or enable all tools for full features:
npx exa-mcp-server tools=web_search_exa,deep_search_exa,get_code_context_exa,crawling_exa,company_research_exa,linkedin_search_exa,deep_researcher_start,deep_researcher_check
Step 4: Use Exa MCP Remotely (Optional)
If you don't want to run Exa MCP locally, you can connect directly to the hosted Exa MCP server using this URL:
https://mcp.exa.ai/mcp
For example, for Cursor and Claude Code, use this HTTP config:
{
"mcpServers": {
"exa": {
"type": "http",
"url": "https://mcp.exa.ai/mcp",
"headers": {}
}
}
}
You can also add your API key in the URL like this for more access:
https://mcp.exa.ai/mcp?exaApiKey=YOUR_API_KEY
Summary
- Install Exa MCP with
npm install -g exa-mcp-serveror run withnpx exa-mcp-server - Add your Exa API key to the configuration for your AI or coding tool.
- Enable the tools you need by adding the
tools=parameter. - Optionally, connect to the remote Exa MCP server using the URL
https://mcp.exa.ai/mcp.
With these simple steps, you can start using Exa MCP to get live, token-efficient web and code context directly inside your AI assistant or coding environment!