AUTOMATIONSWITCH
VendorSearch

Exa Search MCP

by Exa Labs

Neural search across the open web, companies, news, people (LinkedIn), and academic papers, plus a deep-research mode that runs multi-step query plans. 4,371 GitHub stars and 83 commits on main in the last 30 days.

4,371·8 tools·Released JAN 2025·MIT
npx -y exa-mcp-server
Share:

Neural search across the open web, companies, news, people (LinkedIn), and academic papers, plus a deep-research mode that runs multi-step query plans. 4,371 GitHub stars and 83 commits on main in the last 30 days, top cadence in the search category. Exa is the highest-quality general-purpose AI search MCP available at time of review.

Reviewed by M. Nouriel · MAY 2026

INSTALL THIS SERVER

Requires authenticationEXA_API_KEY environment variable. Token from https://dashboard.exa.ai. Token isolation supported: separate keys per tool category for cost attribution.
{ "mcpServers": { "exa": { "command": "npx", "args": [ "-y", "exa-mcp-server" ], "env": { "EXA_API_KEY": "<your-exa-key>" } } } }
PrereqRequires EXA_API_KEY from https://dashboard.exa.ai. NPM: `exa-mcp-server`. For token isolation in production, run multiple instances with distinct keys per tool category. Path: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS).
{ "mcpServers": { "exa": { "command": "npx", "args": [ "-y", "exa-mcp-server" ], "env": { "EXA_API_KEY": "<your-exa-key>" } } } }
{ "mcpServers": { "exa": { "command": "npx", "args": [ "-y", "exa-mcp-server" ], "env": { "EXA_API_KEY": "<your-exa-key>" } } } }
{ "mcpServers": { "exa": { "command": "npx", "args": [ "-y", "exa-mcp-server" ], "env": { "EXA_API_KEY": "<your-exa-key>" } } } }
{ "mcpServers": { "exa": { "command": "npx", "args": [ "-y", "exa-mcp-server" ], "env": { "EXA_API_KEY": "<your-exa-key>" } } } }

8 TOOLS AVAILABLE

web_search_exa
Neural web search with category filters
Read
company_research_exa
Company homepage discovery with metadata (headcount, funding, revenue)
Read
news_search_exa
News coverage search
Read
people_search_exa
LinkedIn profile search (public data)
Read
linkedin_search_exa
LinkedIn-specific search
Read
deep_researcher_start
Launch an asynchronous deep-research task
Write

OUR ASSESSMENT

Strengths
  • 4,371 GitHub stars and MIT licence.
  • 83 commits on main in the last 30 days.
  • Category filters (company, news, people) improve relevance over generic web search.
  • Deep-research mode runs multi-step query plans asynchronously; agent polls for results.
  • Tool restriction via --tools flag for scoped agent configurations.
  • Token isolation: separate API keys per tool category for cost attribution.
  • Browser fallback for sources that require JS rendering.
Weaknesses
  • Exa API pricing is per-request; deep-researcher tasks can be expensive.
  • LinkedIn data quality depends on Exa public scraping; private profiles are unavailable.
  • Category-specific filter restrictions: some filters like includeDomains are scoped to certain categories.
Security Notes

EXA_API_KEY grants full Exa account access. Use token isolation for production: separate keys for web search (cheap) and deep-research (expensive) for cost attribution and rate-limit independence.

Best For

Agents that need neural search with category-aware filters; research workflows where the agent kicks off a deep-research task and polls for completion; coding agents that need source context via get_code_context_exa; teams running production agents who want per-tool API key isolation for cost attribution.

TECHNICAL DETAILS

Language
typescript
Transport
stdio
Clients
Claude DesktopClaude CodeCursorVS CodeWindsurf
License
MIT
GitHub
npm
exa-mcp-server
Last Release
exa-mcp-server (npm latest)MAY 3, 2026
First Released
JAN 1, 2025

ADOPTION METRICS

// GitHub Stars
4,371

// Reading this4,371 stars on the exa-labs/exa-mcp-server repo. 83 commits on main in the last 30 days lead the search category.

// Popularity Rank
#1
Globally · #1 in Search

// Reading thisFirst-ranked in search on commit cadence, star count, and tool surface combined.

SOURCES & VERIFICATION

We don't take any single directory's word for it. Before scoring, we cross-reference 5 public MCP sources, install the server ourselves against the clients we cover, and record when we last re-verified.

01
Discovered
Manual submission
First indexed MAY 3, 2026
02
Cross-referenced
5 directories
PulseMCP, MCP.so, Glama, Smithery, Official MCP Registry
03
Verified against
Claude Desktop, Cursor, VS Code
Installed and tested across clients
04
Last re-checked
MAY 3, 2026
Weekly re-verification
// How other directories see it

The same server, 5 different lenses. We reconcile these signals into our editorial score, which is why our number sometimes diverges from a directory-aggregate star count.

SourceTheir ratingTheir star countTheir downloadsLast synced
AutomationSwitch This page4.6editorial4,371MAY 3, 2026
PulseMCP— unratedunavailableunavailableMAY 3, 2026
MCP.so— unratedunavailableunavailableMAY 3, 2026
Glama— unratedunavailableunavailableMAY 3, 2026
Smithery— unratedunavailableunavailableMAY 3, 2026
Official MCP Registry— unratedunavailableunavailableMAY 3, 2026

// Counts are directory-reported; we don't adjust them. Discrepancies usually come from different snapshot times or star-caching.

// Get in touch

DISCUSS YOUR
MCP REQUIREMENTS.

Evaluating a server, scoping an internal deployment, or working out whether MCP is the right fit at all. Start the conversation and we will point you at the right piece of the ecosystem.

Discuss Your MCP Requirements →