Custodyn

什么是 Custodyn？

AI agents can send emails, execute shell commands, access production systems, and call APIs—but most teams have no runtime control once an agent starts acting. Custodyn is the runtime trust layer for AI agents, enforcing policies before execution with allow/block rules, human approvals, tamper-evident audit trails, and an instant kill switch. Open-core with MIT-licensed Python and JavaScript SDKs, plus a free cloud platform to get started today. Built for OpenClaw, CrewAI, LangChain, MCP etc.