MCPSafe.io
RegistryThreatsMethodologyDocsPricingScanSign in
MCPSafe.io

Security checks for MCP servers — public packages and private repos, fast or deep.

Legal

Privacy PolicyCookie PolicyTerms of ServiceSecurity disclosure

Resources

State of MCP SecuritySupportSystem statusMade in Germany 🇩🇪

© 2026 MCPSafe. All rights reserved.

GDPR — Privacy Policy
← Threat Catalog

Configuration & Environment

Supply chain risks (untrusted servers)

HIGHAIVSS 8.6CWE: CWE-1357OWASP: LLM03Agentic: T13Rule: MCP-100

Installing an MCP server from an unvetted source — typosquat, abandoned package, compromised maintainer — gives the server full capability on your host.

What it is

Supply-chain attacks apply to MCP the same way they apply to npm and PyPI, but with higher stakes: the "installed package" does not just run during your build, it runs any time your agent is active. Typosquats are common; acquisition of abandoned packages by attackers is common; signed releases are not yet common.

Why it matters for MCP

Discovery of new MCP servers today happens through blog posts, Twitter, and curated lists. None of those enforce integrity. A user searching for `mcp-slack` may find three similarly-named packages and no signal about which is canonical.

Vulnerable example

example.js
1
// claude_desktop_config.json
2
{
3
  "mcpServers": {
4
    "slack": {
5
      "command": "npx",
6
      "args": ["-y", "slack-mcp-helper"]   // typosquat of "mcp-slack"
7
    }
8
  }
9
}

Secure example

example.js
1
// Pin by integrity hash and refuse unpinned installs.
2
{
3
  "mcpServers": {
4
    "slack": {
5
      "command": "npx",
6
      "args": [
7
        "-y",
8
        "@modelcontextprotocol/server-slack@1.4.0",
9
        "--integrity=sha512-abc..."
10
      ]
11
    }
12
  }
13
}

How MCPSafe detects this

We compute a Levenshtein distance between new package names and our index of known-good MCP servers, flag suspiciously-close matches, and surface maintainer-change events (ownership transfers) as HIGH evidence.

See the full threat catalog for every documented detection.

Framework alignment

OWASP LLM Top-10 (2025)
LLM03 — Supply Chain
OWASP Agentic AI Top-10
T13 — Rogue Agents
AIVSS v0.5
8.6 (HIGH)AIVSS:1.0/S:HIGH/AV:N/AU:H/BR:H/CD:I

Illustrative CVEs

CVEs of the same CWE class. Not MCP-specific, but exemplify the failure mode MCPSafe detects.

  • CVE-2024-3094 — xz-utils — multi-year social-engineering supply-chain attack

Further reading

  • CWE-1357: Untrustworthy Third-Party Component
  • SLSA framework

Scan an MCP server for this issue

MCPSafe runs this check — and every other rule in the catalog — on any MCP server you paste in.

Scan now