AI Hacking
AI Security Resources
CHECKLIST

MCP Server Pentest Checklist

A comprehensive penetration testing checklist for Model Context Protocol (MCP) servers. Use this to systematically assess MCP server security before deployment.

1 Reconnaissance

  • Identify MCP server endpoints and exposed tools
  • Enumerate available MCP resources, prompts, and tools
  • Map tool parameters and expected input types
  • Check for version disclosure in server metadata
  • Review MCP server configuration files and documentation
  • Identify third-party dependencies and their versions

2 Authentication & Authorization

  • Test for unauthenticated access to MCP tools
  • Attempt authentication bypass via malformed tokens
  • Test for privilege escalation across tools
  • Verify tool-level authorization is enforced
  • Check for IDOR vulnerabilities in resource access
  • Test session/cookie handling and replay attacks

3 Tool Injection Testing

  • Inject malicious parameters into tool calls
  • Test for eval() injection through tool arguments
  • Attempt command injection via tool inputs
  • Test for path traversal in file-accessing tools
  • Verify input length limits and boundary checking
  • Test for SSRF via tool URL parameters
  • Check for NoSQL/SQL injection in tool queries

4 Prompt Injection via MCP Resources

  • Embed injection payloads in MCP resource content
  • Test for indirect injection via third-party data sources
  • Attempt system prompt extraction via tool descriptions
  • Test tool description injection (poisoned tool docs)
  • Verify resource prompt sanitization

5 Data Exposure Testing

  • Check for sensitive data in MCP tool responses
  • Verify data exfiltration through return parameters
  • Test for verbose error messages leaking internals
  • Check logging and monitoring data exposure
  • Test for data leakage through timing side-channels

6 Supply Chain Security

  • Verify MCP server package signatures
  • Review server dependencies for known CVEs
  • Check for backdoors in MCP server code
  • Verify update mechanism security
  • Review tool permissions against least privilege
AH
AI Hacking Team

The AI Hacking team researches and documents AI/LLM security vulnerabilities, red teaming techniques, and defensive strategies. Our guides are based on real-world pentesting experience and continuous monitoring of the AI security landscape.