We Scanned 12 Popular MCP Servers. Here's What We Found.

Date: April 2026 Tool: agent-audit v0.2.1 Method: Static analysis of source code from publicly available MCP server repositories TL;DR We cloned 12 popular, public MCP server repositories from GitHub — including Anthropic's official reference servers — and ran agent-audit's static analysis rules against their source code. We analyzed 1,130 source files across TypeScript, Python, JavaScript, and Go. 58 findings. 12 repos. 100% finding rate. Every single repository had at least one finding. The real concern isn't any individual bug — it's that the MCP ecosystem has no security baseline. There's no equivalent of npm audit for MCP servers, no automated checking in CI, and no community norm around security scanning. Key Numbers Metric Value Repositories scanned 12 Source files analyzed 1,130 Repos with findings 12 (100%) Total findings 58 Critical 12 High 17 Medium 29 Command injection patterns 46 Hardcoded credentials (production code) 7 What We Found 1. Command Injection Is the #1 Risk (4