Detect tool poisoning, over-permissioned access, prompt injection vectors, and supply chain risks in Model Context Protocol servers.
Free security analysis · No signup required · Results in seconds
Malicious instructions hidden in tool descriptions that trick LLMs into harmful actions.
Servers requesting full filesystem, network, or command execution access without justification.
Vulnerabilities allowing attackers to override instructions via crafted tool inputs.
Dependency risks, typosquatting, and compromised packages in the MCP ecosystem.