Security review

Is ai.exa/exa MCP safe to give an agent?

read onlyverifiedverifiedhealthy

A factual risk summary built from ai.exa/exa’s real tool surface, execution model, and verification history — not a vibe. Trust score 84/100.

01What it can do

Only read-style tools observed — no write/destructive tools.

2 tools observed

02Execution model

Runs on the vendor's infrastructure; you connect over the network. No untrusted code runs on your machine, but you grant the hosted service access.

Installs via npx (`npx -y exa-mcp-server`) — it pulls and executes third-party code; pin versions and review the source.

03Permissions & auth

Search the web · write

The write/destructive access this server can exercise, inferred from its verified tools.

04Verification

handshakepassed — tool surface is real

runstools_list:passed · handshake:passed · metadata:passed · tools_list:passed · handshake:passed · metadata:passed · tools_list:passed · handshake:passed · metadata:passed · tools_list:passed

last_checked2026-08-16 19:32Z

sourcesGitHub repo search [p4] · Awesome MCP Servers (wong2) [p6] · Official MCP Registry [p1]

Reduce the risk

Worried about handing an agent raw access? See governed agents in action — Apex gives your AI paced, capped, fully-logged hands with approval queues before anything runs.

Explore Apex →

See also: full server page · setup · alternatives