Claude Code just shipped session recaps, and it's a small feature that solves a genuinely annoying problem.

What it is: When you switch away from a Claude Code terminal session and come back, it now shows a one-line summary of what happened while you were gone.

The problem: Multi-clauding, running several Claude agents in parallel across different terminal windows, means you're constantly context-switching. Coming back to a session mid-task means re-reading output to figure out where things left off.

The solution: After three minutes of inactivity and at least three prior turns, Claude Code generates a recap in the background. It's ready when you refocus the window. You can also trigger it manually with /recap.

What makes it different: The recap generates using prompt cache reuse, so the cost overhead is minimal. It's not a summary you have to ask for, it just shows up.

Key features:

Automatic recap after 3 minutes of inactivity when terminal is unfocused
Manual trigger available via /recap
Configurable off in /config or via environment variable
Works across all plans and providers

Benefits:

Instant reorientation without scrolling back through output
Lower friction when managing parallel AI sessions
Keeps long-running tasks from becoming cognitive overhead

Who it's for: Developers running multiple Claude Code sessions simultaneously, especially those using it for long background tasks like builds, test runs, or agentic workflows.

Honestly, this is the kind of feature that doesn't sound impressive until the moment you actually need it. The use case it targets, multi-clauding across terminal windows, is real and growing.

What's your current setup for managing multiple Claude agents? Curious whether people are using tmux, split terminals, or something else.