# willianpinho/mcp-gateway-scan [Health: Active]

**Category:** 🔒 Security  
**Repository:** https://github.com/willianpinho/mcp-gateway-scan  
**GitHub Stars:** 2  
**npm Downloads (last month):** 152  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/willianpinho-mcp-gateway-scan

## Description
Read-only static scanner for MCP / agent-gateway production-readiness. The scangateway tool scores a repo across 7 security dimensions (RBAC, fail-closed, supply chain, observability, cost, secrets, prod-readiness). Never executes code or prints secret values.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "mcp-gateway-scan": {
    "command": "npx",
    "args": ["-y","mcp-gateway-scan"]
  }
}
```

## Documentation

## What willianpinho/mcp-gateway-scan MCP server does

willianpinho/mcp-gateway-scan MCP server scans source code and configuration for production-readiness anti-patterns in MCP servers and agent gateways. Its MCP interface exposes the `scan_gateway` tool, which accepts a filesystem path and an optional CI setting. The result includes a status for each dimension, severity information, matched evidence, and a structured scan result.

The checks cover seven areas: tool-access governance and RBAC, fail-closed behavior, onboarding and supply-chain pinning, observability, routing and cost controls, secrets and identity, and operational readiness. Findings include examples such as authorization expressed in prompts, error paths that allow requests, unpinned images or packages, missing tracing, absent token or quota limits, inline credentials, and missing kill-switch or evaluation controls.

## How it works

The server reads only the path provided to `scan_gateway`. It does not run the target code, contact external services, or print secret values. When it detects an inline secret, the output identifies the file and line while redacting the value. It also suppresses comments and documentation examples in several checks, helping distinguish executable configuration or prompt content from prose.

Each dimension receives a green, yellow, or red result with a severity tag. The optional `ci` input adds a CI-oriented verdict. The same scanning engine is available outside MCP through the command-line interface, including a `connect` mode that evaluates a repository, MCP client configuration, or package manifest as CONNECT, REVIEW, or DO-NOT-CONNECT.

## Setup and configuration

Install requirements are limited to Node.js 18 or newer. For an MCP client, configure a command equivalent to:

```json
{
  "mcpServers": {
    "gateway-scan": {
      "command": "npx",
      "args": ["-y", "mcp-gateway-scan", "mcp"]
    }
  }
}
```

The package can be added to Claude Code with `claude mcp add gateway-scan -- npx -y mcp-gateway-scan mcp`, or placed in a compatible client's MCP configuration. After setup, ask the agent to scan a repository and have it call `scan_gateway` with a path such as `{ "path": "<repo or dir>", "ci": false }`.

## Tools and capabilities

The primary MCP capability is `scan_gateway`. It can:

- Produce per-dimension green, yellow, and red results.
- Return structured findings with file-and-line evidence.
- Check authorization, error handling, dependency or image pinning, telemetry, budgets, secrets, identity wiring, and operational controls.
- Add a CI gate verdict when `ci` is enabled.
- Preserve read-only behavior and redact secret values.

The command-line mode also supports JSON output, compact CI output, color suppression, and exit codes that distinguish clean scans, red findings, and usage or I/O errors.

## Limitations and notes

willianpinho/mcp-gateway-scan performs static heuristic checks. A green CONNECT verdict means the configured static checks found no red dimensions; it does not verify the publisher or guarantee runtime behavior. The scanner is therefore suited to an initial review and regression gate, not a replacement for a cited security audit or runtime testing.

The scanner only examines the path supplied to it. It cannot assess behavior that is absent from the inspected files, and its conclusions depend on the patterns represented in the repository and configuration. The bundled vulnerable fixtures use fake placeholder secrets for safe testing; those values are not functional credentials.

_Full upstream README: https://allmcps.com/mcp/willianpinho-mcp-gateway-scan/readme_