Back to Home
cisco-ai-defense icon

mcp-scanner

Verified Safe

by cisco-ai-defense

Overview

A Python tool for scanning Model Context Protocol (MCP) servers and tools to detect potential security findings by leveraging Cisco AI Defense API, YARA rules, and LLM-as-a-judge.

Installation

Run Command
mcp-scanner-api --host 0.0.0.0 --port 8080

Environment Variables

  • MCP_SCANNER_API_KEY
  • MCP_SCANNER_ENDPOINT
  • MCP_SCANNER_LLM_API_KEY
  • MCP_SCANNER_LLM_MODEL
  • MCP_SCANNER_LLM_BASE_URL
  • MCP_SCANNER_LLM_API_VERSION
  • MCP_SCANNER_LLM_TIMEOUT
  • AWS_REGION
  • AWS_PROFILE
  • AWS_ACCESS_KEY_ID
  • AWS_SECRET_ACCESS_KEY
  • AWS_SESSION_TOKEN
  • MCP_SCANNER_YARA_RULES_DIR
  • MCP_SCANNER_OAUTH_CLIENT_ID
  • MCP_SCANNER_OAUTH_CLIENT_SECRET
  • MCP_SCANNER_OAUTH_REDIRECT_URI

Security Notes

The server demonstrates good security practices such as using environment variables for API keys, employing Pydantic's SecretStr for sensitive data, and explicit authentication controls. The LLM analyzer uses random delimiters to mitigate prompt injection. The `stdio` server interaction carefully normalizes and validates commands using `shutil.which` and `shlex.split` before execution, which is a strong mitigation for command injection in that context. However, any interaction with external processes, especially in `stdio` mode, inherently carries risk and requires trust in the executed command. The project's documentation explicitly covers security policies and vulnerability disclosure.

Similar Servers

Stats

Interest Score98
Security Score8
Cost ClassHigh
Avg Tokens2000
Stars665
Forks57
Last Update2025-12-06

Tags

SecurityAIMLScanningAPIYARALLMCLIREST APIModel Context Protocol