Do you have visibility into which MCP servers your devs are connecting agents to?
@zachlloydtweets (@warpdotdev) says the real risk isn't the protocol. It's every dev setting up their own tool dependencies with no central oversight:
CISO panel episode on governed tool access:
Docker: Conquer the complexity of app development.
- The dev approved the git branch. The check found nothing wrong. The attack worked anyway. Part 5 of @ajeetsraina's AI Coding Agent Horror Stories series: how a two-line injection slipped past an allowlist that was working exactly as designed: bit.ly/3RUUYck
- Docker Sandboxes is now in the official @AnthropicAI Claude Code docs as a recommended way to run Claude Code with microVM-level isolation. Its own kernel, its own Docker daemon, no Docker Desktop required: bit.ly/4hG0cmH
- New leadership at Docker: @matvelloso joins as Chief Product Officer and Vinh Le as CFO. Both bring deep AI product knowledge and leadership experience as we double down on the agentic era. Welcome aboard! Read more: bit.ly/3U6OTdl


