Hey <!here>! As Jake mentioned above, we've recently launched a new open source secrets management project called Agent Vault that's meant to help AI agents use secrets without letting them read the underlying secrets; this works as an egress proxy that agents are configured to proxy requests through agnostic to the interface that they may be using: API, CLI, SDK, MCP. The pattern that we're seeing more and more teams adopt and build on their own is called credential brokering and it's meant to solve a problem called credential exfiltration where agents can be prompt injected into returning/leaking secrets. We believe that this new infrastructure component will increasingly become the common way that teams securely build and deploy AI agents in a way that agents cannot be manipulated into leaking secrets because they don't have direct access to them in the first place; happy to answer any questions about it as well. If this sounds interesting, I'd highly recommend checking it out and joining the below new channels: #C0B0CMEGQ10 #C0AV4K8LGBD #C0B09DU9LMT Onward and upward 🎉🦠