mcp-watch

MCP Watch is a security scanner specifically designed for MCP (Model Context Protocol) servers. It analyzes MCP implementations for a broad range of vulnerabilities and misconfigurations, including credential leaks, tool poisoning, prompt and parameter injection, data exfiltration, protocol violations, and access-control issues. The tool aggregates findings across categories such as credential-leak, tool-poisoning, data-exfiltration, prompt-injection, tool-mutation, steganographic-attack, protocol-violation, input-validation, server-spoofing, toxic-flow, and access-control, enabling you to prioritize remediation efforts. Use it to scan GitHub-hosted MCP servers or local MCP projects, with options to tailor output and severity filters for focused reporting.

Prerequisites:

• Node.js and npm installed on your system
• Internet access to fetch the MCP Watch package

Install globally (recommended for quick usage):

npm install -g mcp-watch

Install locally (for project-specific usage):

npm install mcp-watch

From source (if you prefer building from repository):

git clone https://github.com/kapilduraphe/mcp-watch.git
cd mcp-watch
npm install
npm run build

Usage examples after installation:

• Scan a GitHub MCP server:

mcp-watch scan https://github.com/user/mcp-server

• Scan a local project:

mcp-watch scan-local /path/to/your/mcp-project

For Docker usage, refer to the Docker installation steps in the README to run as a container, if you prefer containerized scanning.

Notes and tips:

• Use the --format json option to get structured output suitable for further processing.
• You can filter scans by severity (--severity) and category (--category) to focus on high-priority issues.
• Ensure your network allows access to the target GitHub repository if you’re scanning remote sources.
• Keep MCP Watch updated to receive the latest vulnerability checks and categories.
• The tool categories map to common MCP security concerns; if you encounter a new category, consider contributing to the scanner’s rule set.