Claude Code adds an agent view for running many sessions at once

Claude Code has a new agent view: a single screen for managing many sessions in parallel instead of juggling terminal tabs. Each session appears as a row showing its last response, whether it is waiting on you, and how long it has been idle. You open it by pressing the left arrow inside Claude Code or running claude agents. From there you can reply inline to a session, attach to one for the full transcript, and move existing sessions into the background view with /bg. To start a task in the background from the start, use claude --bg [task].

The piece is small but the design choice is worth thinking about. Claude Code could already run long tasks, but the friction of remembering which terminal window owned which task pushed people toward one-at-a-time use. Treating sessions like rows you scan, rather than windows you switch between, fits the way agent work is actually shaped: asynchronous, with the model thinking for minutes at a time. Anthropic writes that early users are running concurrent sessions, watching long-running tasks, and tracking deployments through the view. It ships as a Research Preview across Pro, Max, Team, Enterprise, and the Claude API, under existing rate limits.

There is no API or pricing change here, just a workflow upgrade. The open question is how many parallel sessions are actually useful before the human reviewer becomes the bottleneck.

Why it matters

If you have been running Claude Code one task at a time, agent view is the obvious next step: try claude --bg on background work like refactors or test fixes, and see how many you can supervise at once before the rate limit, or your own attention, becomes the limit.