> MCP-Guard

Security scanner for Model Context Protocol (MCP) servers. Paste a tools/list JSON manifest or Python/TypeScript server code to find hidden attacks in the tools AI agents use, then get AI explanations of each finding.

MCPG-001 tool poisoningMCPG-002 excessive permissions MCPG-003 rug-pullMCPG-004 prompt-injection paths MCPG-005 hard-coded secretsOWASP LLM Top 10 (2025)

Filename (for source)

AI explanations (OpenAI)

Static analysis only: submitted code is never executed. Secrets are redacted before anything is sent to the AI model. Do not paste real credentials. The samples use saved AI reviews; AI on your own input is limited to a few scans per hour.