Vulnify

Vulnify is runtime authorization for AI agents. Before a sensitive tool call runs, your code calls check() and gets ALLOW, REVIEW or BLOCK, with a 0-100 risk score and the reasons: action, data sensitivity, destination, volume, time of day. REVIEW pauses only that action until a person approves. Start in monitor mode, fail closed by default. Policies as code, Node and Python SDKs (MIT) with LangChain, MCP, CrewAI and LangGraph adapters, and a hash-chained audit log with JSON/CSV export. Free plan available.