How to Install Exa MCP in Claude Code
Neural search for AI agents — search by meaning, find similar content, and get research-grade web results
What Exa MCP does
Exa is the search API designed from the ground up for AI and semantic search. Unlike traditional keyword search, Exa understands meaning — finding conceptually related content even without matching keywords. Features: - Neural search — finds content by meaning, not just keywords - Find similar pages to any URL - Research mode — comprehensive, multi-source research synthesis - Date-filtered search for recent content - Domain-specific search (academic papers, news, GitHub, etc.) - Structured content extraction with highlighted passages - Competitor research and landscape analysis - Technical documentation search Where Exa wins over Brave/Tavily: - "Find papers similar to this arXiv abstract" — understands semantic similarity - "Find companies doing what Figma does but for 3D" — finds conceptual matches - "What was written about [topic] before 2023?" — precise temporal filtering Exa is the choice for research-intensive AI workflows where query precision matters.
Installation steps
- 1Copy the config below
- 2Open or create .claude/mcp.json in your project root
- 3Paste inside "mcpServers": {}
- 4Reload Claude Code
Configuration
File location: .claude/mcp.json (project root) or ~/.claude/mcp.json (global)
{
"exaMcp": {
"command": "npx",
"args": [
"-y",
"exa-mcp-server"
],
"env": {
"EXA_API_KEY": "your_exa_api_key"
}
}
}Install in other clients
For maintainers
If you maintain Exa MCP, add this badge to your README to show it's verified on CuratedMCP:
[](https://curatedmcp.com/marketplace/exa-mcp)
Related servers
Brave Search MCP
Real-time web and local search using the Brave Search API — no tracking, no ads
Perplexity MCP
AI-powered search with citations — real-time web answers grounded in up-to-date sources
Tavily MCP
AI-optimised web search built for agents — grounded, real-time answers with source citations