Skills/npm/brave-search-mcp-server

brave-search-mcp-server@2.2.0

VERIFIED

Publisher claims: "Brave Search MCP Server: web results, images, videos, rich results, AI summaries, and more." Observed in sandbox: no network, touches nothing outside the project, writes no memory.

Registry npm (MCP server)Publisher brave · publishing for 491 daysInstalls 197KLast detonated 7 h ago
$claude mcp add brave-search-mcp-server -- npx -y @brave/brave-search-mcp-server
WatchReport an issue

Coverage

Started with a decoy BRAVE_API_KEY; tool list observed. Tool calls with a real credential not tested.
Tested
server startup with a tagged decoy in BRAVE_API_KEY · MCP initialize and tools/list · every tool description, for hidden instructions · where the decoy credential went · every file read, file written and host contacted by what ran
Not tested
what the tools do when called with a real credential
Needs
BRAVE_API_KEY — give it to the server where it runs, through vet-skills run --secrets, so the value never enters the server's environment. Never paste a credential into a chat, a ticket or a report.

Before you give it your token

It will ask for
API_KEYBRAVE_API_KEYBRAVE_API_KEY_FILE or --brave-api-key / --brave-api-key-fileConfiguration, not credentials: BRAVE_MCP_ALLOWED_HOSTS, BRAVE_MCP_ALLOWED_ORIGINS, BRAVE_MCP_DISABLED_TOOLS, BRAVE_MCP_ENABLED_TOOLS, BRAVE_MCP_HOST, BRAVE_MCP_LOG_LEVEL, BRAVE_MCP_PORT, BRAVE_MCP_STATELESS, BRAVE_MCP_TEST_IMAGE, BRAVE_MCP_TRANSPORT, CHAT_COMPLETIONS_PATH
Where it talks to
brave.com (its own)Also slack.com, a.example, b.example, c.example, allowed.example, evil.example, amazon.com, a.com, b.com, reuters.com, nytimes.com, bbc.com, civitatis.com, britannica.com, visualcapitalist.com, cntraveler.com — not the publisher's domain; check what for before connecting.
What we saw when it ran
no credential files read · no network
What we could not test
what the tools do when called with a real credential

Connect it with a scoped credential, under the guard, and read the third-party hosts first. The code names slack.com, a.example, b.example, c.example, allowed.example, evil.example, amazon.com, a.com, b.com, reuters.com, nytimes.com, bbc.com, civitatis.com, britannica.com, visualcapitalist.com, cntraveler.com as well as its own vendor. The guard injects API_KEY, BRAVE_API_KEY, BRAVE_API_KEY_FILE only into requests to brave.com (and its subdomains) and refuses the rest.

$vet-skills run --secrets API_KEY=brave.com,BRAVE_API_KEY=brave.com,BRAVE_API_KEY_FILE=brave.com -- npx -y @brave/brave-search-mcp-server

Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.

What changed since 2.1.5

vet-skills diff brave-search-mcp-server 2.1.5 2.2.0

This version reaches for something 2.1.5 did not: 6 new hosts (a.com, b.com, britannica.com, civitatis.com, …) · 1 new tool (brave_answers) · 3 rewritten tool descriptors (brave_local_search, brave_summarizer, brave_web_search). Compared from the two recorded runs, not from source.

New hosts
a.com, b.com, britannica.com, civitatis.com, cntraveler.com, visualcapitalist.com
New tools
brave_answers
Rewritten tool descriptors
brave_local_search, brave_summarizer, brave_web_search

The guard can hold an update like this until someone approves it (runtime.on_change: hold). A change that only shows with a real credential, or on a path the run did not reach, is not in a trace and so not in this comparison.

Is [email protected] safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios) on 2026-11-19 and scored it 91/100 (Verified). It did what it claims and nothing else we could detect: no undeclared network egress, no credential reads, no persistent instruction writes. Summary: no network, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 46 s · 2 scenarios · 0 deviations
t+CategoryObserved action
0.0sprocessno install scripts in bundle
5.9sprocessdependency install exit 0
26sprocessrelaunched with a decoy BRAVE_API_KEY (tagged; nothing real)
46sclaimed9 tools with a decoy BRAVE_API_KEY: brave_web_search, brave_local_search, brave_video_search, brave_image_search, brave_news_search, brave_summarizer, brave_llm_context, brave_place_search, brave_answers
46stoolse1dab301b21a57f7
46stools9515cbcabd492888
46stools70cd77f52a959c9d
46stoolsb603b19d20d0342c
46stools6138e22a4f20a920
46stools9e8664a3c3e0e8d5
46stools5a904701f573c4e6
46stools824a9ba00b7126e1
46stools50bea30479e2f723
46sprocessscript backend runs setup only
46sprocessno install scripts to re-run

Static findings

HIGH
Access to sensitive file pathsdist/docker.e2e.js:40 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system filesenv: { ...process.env, ...env },Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05

No other findings across 6,006 lines in 75 files.

Cost to the agent

Instructions loaded
≈66 tok
1 file · 75 in bundle
Tool list per prompt
≈9,496 tok
9 tools · 1055 per tool
Handshake
3291 ms
launch to tools/list, stdio
Undocumented tools
0
description under 20 chars · avg 866

Tokens estimated at four characters each. The tool list is re-read by the agent on every turn, so it is the number to compare between servers that do the same job: ranked across the index.

Known vulnerabilities

No advisories in OSV for @brave/brave-search-mcp-server@2.2.0 as of 6 h ago. This is the package itself, not its dependencies.