Is Exa MCP server safe?

What Exa MCP server does

Exa MCP Server connects AI agents to Exa for web search, page content fetching and multi-step research, through tools such as web_search_exa, web_fetch_exa and agent_run. It is published in the exa-labs/exa-mcp-server repository and is used mainly through Exa's hosted endpoint at mcp.exa.ai. It sends your queries and URLs to Exa's service; it works anonymously with rate limits, and OAuth or an API key unlocks higher limits and Exa Agent. Before installing, prefer OAuth or an Authorization header over putting the API key in the URL with ?exaApiKey=, and use the tools parameter to enable only what you need.

Low risk - review the findings before installing

Notable — review in context (a capability is not a malicious indicator):

Server bound to all network interfaces
Node.js network egress
npm prepare hook

What to do: Nothing here argues against installing it. Grant the capabilities it lists only if you expect the tool to need them.

Exa MCP server is an MCP server analyzed by SkillTotal's deterministic static scanner. The scan found no malicious indicators, though 1 risky construct is reported for review. It can: install time execution, mcp tools detected and network egress — capabilities are what the code can do, not a verdict on intent. Risk score 10/100 (low).

exa-mcp-server 3.4.3

GitHub repository, read as an MCP server · https://github.com/exa-labs/exa-mcp-server

LOW

/ 100 risk score

Snapshot · scanned Nov 13, 2026 · exa-mcp-server@3.4.3 · engine 1.1.0 / ruleset 100

Get notified when this component changes

Change alerts aren't live yet. Leave your email and we'll let you know when they launch. We use it only for this and delete it on request.

Automated static-analysis result. It can contain false positives and false negatives, and is not a claim about the intent of Exa MCP server's authors. Report a false positive.

Capabilities — what this component can do (not a risk score):

install time executionmcp tools detectednetwork egress

Behavioral traits

How this component maps to the CSA agentic threat model. Descriptive — it never affects the risk score.

Tool surface

Tool Usage

Network egress

Interaction & Communication / Direct Communication

Network exposure

Interaction & Communication

Supply-chain provenance risk

General Protections / Supply Chain

Findings (4)

MEDIUMServer bound to all network interfacesST-EXPOSE-BIND

A server is bound to all network interfaces (0.0.0.0), not just your own machine.

src/runtime-server.ts:5

const HOST = process.env.HOST || "0.0.0.0";

Why it matters: Without authentication, other hosts on the network can reach it.

Fix: Bind to 127.0.0.1 for local-only use, or require authentication and restrict access if remote exposure is intended.

Capabilitynpm prepare hooknot scored · severity mediumST-INSTALL-NPM-PREPARE

package.json has a 'prepare' script (runs on git/local installs and before publishing).

package.json:52

"prepare": "npm run build:stdio",

Why it matters: Usually a build step, but confirm it doesn't fetch or run remote code.

Fix: Usually a legitimate build step; confirm it only builds and does not fetch or execute remote code.

CapabilityNode.js network egressnot scored · severity mediumST-NET-NODE

The component makes outbound network requests.

src/runtime-server.ts:1

import { createServer, type IncomingMessage, type ServerResponse } from "node:http";

Why it matters: Usually legitimate, but confirm the destinations are expected and no sensitive data leaves.

Fix: Confirm the destination hosts are expected and that no sensitive data is sent off-host.

CapabilityMCP tool surface detectednot scored · severity lowST-MCP-DETECTED

An MCP tool surface (manifest or tool definitions) was found.

.claude-plugin/plugin.json:10

"mcpServers": {

gemini-extension.json:5

"mcpServers": {

mcp.json:3

"mcpServers": {

mcp.json:1

plugins/exa/.claude-plugin/plugin.json:14

"mcpServers": {

server.json:1

src/stdio.ts:45

const server = new McpServer({

src/stdio.ts:64

await server.connect(new StdioServerTransport());

src/tools/agentRun.ts:674

server.registerTool(

src/tools/companyResearch.ts:16-17

server.tool(
    "company_research_exa",

src/tools/deepResearchCheck.ts:19-20

server.tool(
    "deep_researcher_check",

src/tools/deepResearchStart.ts:14-15

server.tool(
    "deep_researcher_start",

src/tools/deepSearch.ts:29-30

server.tool(
    "deep_search_exa",

src/tools/exaCode.ts:16-17

server.tool(
    "get_code_context_exa",

src/tools/linkedInSearch.ts:16-17

server.tool(
    "linkedin_search_exa",

src/tools/peopleSearch.ts:16-17

server.tool(
    "people_search_exa",

src/tools/webFetch.ts:39

server.registerTool(

src/tools/webSearch.ts:181

server.registerTool(

src/tools/webSearchAdvanced.ts:147-148

server.tool(
    "web_search_advanced_exa",

Why it matters: Just context — review which tools it offers and their permissions.

Fix: Review the declared MCP tools and their permissions.

How attackers abuse these capabilities

Interactive labs on the attack class behind the rules above. They show the technique, not anything found in Exa MCP server.

Install Script lab

Maintain Exa MCP server?

Check the next package before your agent installs it

The free Claude Code plugin scans each npm, PyPI or MCP package with this report's open-source engine before the install command runs. It runs on your machine and needs no account.

Set up the plugin Or scan pull requests in CI with the GitHub Action

Check your own component

Run the same evidence-backed scan on any MCP server, agent skill, or package.

Scan your own component

How we determine this: deterministic static analysis (regex + AST), evidence-anchored, no code execution. Methodology →