Anthropic adds "agent view" to Claude Code

Anthropic adds agent view to Claude Code, letting users manage multiple active terminal sessions and background tasks in a single consolidated interface.

Claude Code integrates a new interface, agent view, which consolidates all active terminal sessions in one place. Until now, running multiple agents in parallel required juggling between tabs, tmux panes, and mentally tracking tasks to resume.

The screen lists all sessions line by line, along with their status (in progress, completed, awaiting a decision), the content of their last response, and the timestamp of the last interaction. It can be accessed from any session via the left arrow, or directly with the `claude agents` command. Selecting a line allows you to preview the last exchange, reply inline if validation is expected, or attach the session to explore the full transcript. The `/bg` command sends an existing session to the aggregated view, while `claude --bg [task]` directly opens a session in the background without foregrounding it.

Several use cases have emerged from early users: multiplication of concurrent sessions (each potentially linked to a skill), supervision of long-running agents such as "PR babysitters" or dashboard updates, and quick navigation between an ongoing task and a specific question about the codebase.

Agent view is available as a Research Preview on Pro, Max, Team, Enterprise, and API plans, within the usual limits of each plan.