Exa MCP
by Exa
Gives an agent real, current web search built on Exa's AI-native search API — designed for retrieving high-quality sources rather than raw SERP scraping.
Add to Claude Code
claude mcp add exa -- npx -y exa-mcp-server
About
Exa MCP connects an agent to Exa's search API, which is built specifically for AI use cases: semantic, embeddings-based search that returns full page content and highlights rather than just links and snippets.
Beyond general web search, Exa's API includes specialized modes for company research and academic paper search, which the MCP server exposes as separate, purpose-built tools.
Key features
- Semantic web search built for AI retrieval, not just link lists
- Returns full page content and highlights, not just snippets
- Specialized tools for company research and academic paper search
- Simple API-key-based setup, no browser automation required
Use cases
- Give an agent real-time web search to answer questions beyond its training data
- Research a company's background before drafting outreach or an analysis
- Pull recent academic papers on a topic directly into a research workflow
Available tools
web_search_exa
Performs a semantic web search and returns page content for top results.
company_research
Searches for and compiles information about a specific company.
research_paper_search
Searches academic papers and returns relevant results with content.
Frequently asked questions
Do I need an Exa API key?
Yes — set EXA_API_KEY as an environment variable; a key is available from the Exa dashboard, including a free tier for evaluation.
How is this different from a generic web search MCP?
Exa's underlying API uses semantic embeddings rather than keyword matching, generally surfacing more relevant sources for research-style questions.