Anthropic just enhanced Claude Code with sub-agents. They pair nicely with Google Cloud Security’s Model Context Protocol (MCP) servers and I’ve updated my SOC AI Runbooks, so that .claude/agents/ use the existing SOC Personas.

Each sub-agent executes its security operations task with context isolation. That more focused context leads to more efficient, cost-effective, and faster execution. Their findings are then returned for synthesis without, as far as I can tell, degradation to the final artifacts. I’m going to add this work as a submission to the MCP Community Challenge and we’d like to see yours too: “This is your chance to contribute to the Community, show off your skills, inspire others and win some awesome Google swag!”

In the video below, check out my first attempts at using the Claude Code sub-agents with the prompt: “use the available subagents to investigate case 3623”. I was impressed with the final report, so be sure to watch (or skip) to the end…and don’t forget to like and subscribe! ;-)