@mlawsonking/code-guard-mcp

MCP server: security scanner for AI-generated code — the agent scans its own code/diff before committing. Detects injection, SSRF, hardcoded secrets, weak crypto, unsafe deserialization, TLS-off, XSS. Deterministic, free, no LLM.

Source

npm

Version: 1.0.0

Version published: 2 months ago

Weekly downloads: 61

Maintainers: 1

Weekly downloads

Created: 2 months ago

Source

code-guard-mcp

Security scanner for AI-generated code, as an MCP server — the check a coding agent runs on its own code or diff before it commits. Deterministic, free, no LLM.

Why: 53% of code is now AI-written and ~25% ships vulnerable, yet nothing scans it in the agent's loop for free. Code Guard is that fast first-line scanner (not a full audit replacement).

Tools

scan_code — scan a snippet → findings {rule, category, severity, line, message, remediation} + verdict pass/review/block. Detects command/code/SQL injection, SSRF, hardcoded secrets, weak crypto, unsafe deserialization (pickle/yaml), disabled TLS verification, XSS / template injection.
scan_diff — scan only the added lines of a unified diff (correct new-file line numbers).
list_rules — the rule catalog (coverage transparency).

Install

{ "mcpServers": { "code-guard": { "command": "npx", "args": ["-y", "@mlawsonking/code-guard-mcp"] } } }

Deterministic (same input → same output). API: https://code-guard-api.vercel.app · part of the agent-guardrail suite (Package Guard · Agent Firewall · Payment Guard · Email Guard · Code Guard).