Original article in Spanish on El Rack: Claude Code ya no espera a que le mires: subagentes en segundo plano por defecto
I run Claude Code daily as my main dev agent, both locally and over SSH against a VPS running a handful of production sites. A change that landed over the last few weeks quietly rewired how I work with it: subagents now run in the background by default.
Before, when Code delegated a task to a subagent, the conversation blocked until that subagent finished. Since July 1, it keeps working on other things while the subagent runs, and notifies you when it's done. For anyone running long sessions with --dangerously-skip-permissions and checking back hours later, that's the difference between a tool that waits and one that actually works unattended.
What changes day to day
My workflow has always been the same: approve the initial plan, then let Code proceed without confirmation on every step, stopping only on real errors. With background subagents, that autonomy compounds. I can now ask it to audit a site, generate new content, and review config files at the same time — each running as an independent subagent instead of queuing one after another.
