Goop Shield

소개

Goop Shield offers robust runtime defense for AI agents, safeguarding them against a spectrum of adversarial attacks including prompt injection, data exfiltration, and configuration tampering. It operates through a comprehensive pipeline of 24 inline defenses and 3 output scanners, deployable as an HTTP API server, a first-class Model Context Protocol (MCP) server, or a Python SDK, ensuring your AI systems remain secure and operate as intended with full audit and telemetry capabilities.

주요 기능

• 24 Inline Defenses for prompt injection, exfiltration, config guarding, and more
• 3 Output Scanners for secret, canary, and harmful content leak detection
• Red Team Validation with a built-in adversarial probe framework
• MCP Server supporting Claude Code, Cursor, Windsurf, and other AI agents
• Framework Adapters for seamless integration with LangChain, CrewAI, and OpenClaw
• 1 GitHub stars

사용 사례

• Providing runtime security for AI coding agents integrated via Model Context Protocol (MCP).
• Protecting AI agents from prompt injection, data exfiltration, and configuration tampering.
• Securing large language model applications built with frameworks like LangChain and CrewAI.