Jörg Michno Shield Blog Audit Registry GitHub

← Blog · ClawGuard Shield

We Scanned 4,113 MCP Servers from the Official Registry — Here's What We Found

By Joerg Michno · March 28, 2026 · 5 min read

1.6%
of 4,113 MCP servers flagged with security-relevant patterns

We scraped the entire MCP registry (11,529 entries), resolved 4,113 unique servers with GitHub repos, and ran static pattern analysis on their tool descriptions, READMEs, and registry metadata. 65 servers were flagged. Here's what we found — and what it means.

Methodology

11,529

Registry entries scraped from registry.smithery.ai, mcpservers.org, and glama.ai (aggregated)

4,113

Unique GitHub repos after deduplication

223

Detection patterns across 15 languages

4.1s

Total scan duration for all 4,113 servers

We used ClawGuard for static analysis: 225 patterns, 15 languages, 10-stage preprocessing including Unicode normalization, leetspeak decoding, homoglyph detection, Base64 decoding, ROT13, and delimiter stripping.

Important: This is pattern matching on text content — tool descriptions, READMEs, and registry metadata. This is not behavioral probing or dynamic testing. We're looking at what servers say they do, not what they actually do at runtime.

Results by Severity

SeverityCountExamples
CRITICAL7Invisible Unicode, private keys, curl pipe bash
HIGH38SSRF-capable tools, code execution patterns, data dump capabilities
MEDIUM24Connection strings (often examples), broad file access
LOW12Overly permissive tool descriptions

Total: 81 findings across 65 servers. Some servers triggered multiple patterns.

Top Findings by Category

SSRF / Fetch Risk
18 servers

Tools with webfetch capabilities that can request arbitrary URLs. In an agent context, this means the AI can be instructed to fetch internal endpoints, cloud metadata APIs, or exfiltration URLs.

Code Execution
17 servers

curl|sh install patterns and exec capabilities in tool descriptions. An AI agent following these instructions would execute arbitrary shell commands without human review.

Browser Automation
12 servers

DevTools and Playwright-based servers with unrestricted page access. These tools can navigate to any URL, execute JavaScript, and extract page content — a powerful exfiltration vector.

Credential Exposure
11 servers

Connection strings, API keys, and tokens found in READMEs and tool descriptions. Most are example strings in documentation — but automated tools and agents don't distinguish examples from real credentials.

Data Exfiltration
6 servers

Webhook URLs and tool descriptions that explicitly mention "dump all data" capabilities. Combined with an LLM that follows instructions literally, this is a direct data loss vector.

Notable Findings

1. Invisible Unicode (china-stock-mcp)

Zero-width joiner characters found in tool descriptions. Whether this is an encoding artifact or intentional, it's exactly how prompt injection attacks hide instructions from human reviewers. The text looks normal to a person, but the model processes hidden tokens that can alter behavior.

2. curl|sh Install Patterns

Multiple servers instruct users to pipe curl output directly to sh. This is already risky for humans — in an MCP context, an AI agent could execute this without any human review step, giving the remote server arbitrary code execution on the host machine.

3. Private Key in README (apple-ads-mcp)

A BEGIN PRIVATE KEY block was found in documentation. Likely a template placeholder, but automated tools and AI agents don't distinguish placeholders from real secrets. An agent ingesting this README could leak the key in a response or log.

4. Data Exfiltration Descriptions (session-forge)

Tool descriptions that explicitly mention "dump all data" capabilities. When an LLM reads a tool description that says "dump all data to webhook," it treats that as a legitimate capability — and may invoke it when a user asks to "export everything."

Important Caveats

Want to check your own MCP server? Generate a free EU AI Act compliance report — maps findings to Art. 9, 13, 15, 61.

Generate your free compliance report →

What This Means for Teams Using MCP

Audit tool descriptions before connecting

Before you connect an MCP server to a production agent, read its tool descriptions. If a tool says it can fetch arbitrary URLs or execute commands, decide whether your agent actually needs that capability.

Use input scanning as defense-in-depth

Run a scanner like ClawGuard on tool descriptions and agent inputs. Static analysis won't catch everything, but it catches the low-hanging fruit — hidden Unicode, known injection patterns, credential leaks.

Be cautious with fetch/exec capabilities

Servers with webfetch, exec, or browser automation tools are high-privilege. Treat them like you'd treat a third-party npm package with postinstall scripts — review before you trust.

EU AI Act compliance is concrete

Article 15 requires "appropriate levels of accuracy, robustness, and cybersecurity." Scanning your MCP stack and documenting the results is a concrete compliance step — and the August 2026 deadline is approaching.

Scan your MCP server inputs for free — no account needed.

Try the free scan →

View the scanner on GitHub (MIT licensed) →