home / mcp / exa mcp server

Exa MCP Server

Connects AI assistants to Exa AI search, research tools, and code context for live web results and up-to-date code examples.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "agentwings-exa-mcp-server": {
      "url": "https://mcp.exa.ai/mcp",
      "headers": {
        "EXA_API_KEY": "REPLACE_WITH_YOUR_API_KEY"
      }
    }
  }
}

You can connect AI assistants to Exa AI’s smart search and code-context capabilities through an MCP server. This server enables real-time web search, code-context retrieval, and deep research tooling so your agents can answer questions with live data and relevant code examples.

How to use

Use an MCP client to connect to Exa’s MCP server to access web search, code context, and other research tools. You can run the server locally with a standard runtime like NPX, or connect remotely via the HTTP endpoint. Once connected, you can enable specific tools to tailor the assistant’s probing and retrieval behavior, such as code search for up-to-date examples or web search for live results.

How to install

Prerequisites: Node.js and npm should be installed on your machine. You can verify installations with node -v and npm -v.

Option A: Run the MCP server remotely (HTTP endpoint). This uses a remote MCP URL that you connect to from your client.

Option B: Run the MCP server locally with NPX. This starts the server directly in your environment.

Follow the steps below to set up the HTTP-based remote server configuration and, if desired, a local NPX-based startup.

1) HTTP remote configuration (recommended for remote hosting):

{
  "mcpServers": {
    "exa": {
      "type": "http",
      "url": "https://mcp.exa.ai/mcp",
      "headers": {}
    }
  }
}

Additional configuration steps you may need

To connect Claude Desktop or Cursor/Claude Code to the remote MCP, you can add the following HTTP-based configuration. This enables the integration without running a local server.

{
  "mcpServers": {
    "exa": {
      "type": "http",
      "url": "https://mcp.exa.ai/mcp",
      "headers": {}
    }
  }
}

Using NPX to run the Exa MCP server locally

If you prefer to run the server locally, you can start it with NPX. This runs the MCP server and enables you to specify tools if needed.

npx exa-mcp-server
```

You can enable specific tools or all tools by passing a comma-separated list after the server binary, for example:

```
npx exa-mcp-server tools=web_search_exa,deep_search_exa,get_code_context_exa
```

Or enable all tools:

```
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

Available tools

get_code_context_exa

Search and retrieve relevant code snippets, examples, and documentation from open-source libraries, GitHub repositories, and programming frameworks. Useful for finding live-verified usage patterns and API examples.

web_search_exa

Perform real-time web searches with optimized results and content extraction to provide current information.

deep_search_exa

Deep web search with smart query expansion and high-quality summaries for each result.

company_research_exa

Crawl company websites to gather detailed business information for companies.

crawling_exa

Extract content from specific URLs, enabling reading of articles, PDFs, or any web page when you have the exact URL.

linkedin_search_exa

Search LinkedIn for companies and people using Exa AI, with optional inputs like company names or URLs.

deep_researcher_start

Start a smart AI researcher for complex questions that gathers web sources and produces a detailed research report.

deep_researcher_check

Check the progress and retrieve the comprehensive research results.