Claude Code Agents: A Dashboard for Your AI-Powered Terminal Sessions

If you run multiple Claude Code sessions at once -- juggling a perf audit in one terminal, drafting release notes in another, and generating landing page copy in a third -- you know the pain of tab-switching. Anthropic's new Agent View consolidates all of your Claude Code sessions into a single, scannable dashboard. Chase AI walks through exactly how it works and why it matters for power users running parallel AI workloads.

The Problem: Too Many Terminals

Anyone running Claude Code seriously ends up with three, four, five, or more terminal windows open simultaneously, each working on a different task. The real cost is not screen real estate -- it is cognitive overhead. You switch to one session, get pulled into something else, and forget about another session for an hour and a half. If you do not visually see it, you tend to forget about it.

Title screen showing "agent view in Claude Code" with the Claude mascot and terminal windows in the background

Launching Agent View

Getting started is straightforward. Type claude agents in any terminal and a dashboard appears, organized into three sections: Needs input, Working, and Completed. The header shows Claude Code version info, the model (Opus 4.7 with 1M context), and a summary of session counts.

Claude Code Agent View dashboard showing "Needs input", "Working", and "Completed" sections with multiple named sessions

The Dashboard Layout

The dashboard in Chase's demo shows real sessions at various stages. Under "Needs input," sessions like "dark-mode" and "release-notes" are flagged and waiting for a human decision. Under "Working," active sessions like "perf-audit," "payment-migration," "onboarding-copy," and "load-test" churn away autonomously. Under "Completed," finished work like "test-coverage" shows its result -- billing coverage taken from 61% to 92%, with a PR already filed.

Navigating and Interacting

Clicking on any session opens the full terminal view, just as if that terminal were open on its own. Press the left arrow to return to the dashboard. But the real productivity feature is the quick-peek: mouse over a session and hit spacebar to see elapsed time, a status summary, and a reply box. It is like poking your head into a room -- see what is going on, give a quick instruction, and move on without losing your place in the dashboard.

Claude Code agents dashboard showing live sessions with status summaries, elapsed times, and a prompt to describe a new task

Peeking and Replying Inline

The peek view shows a condensed summary of what the session has accomplished so far. In the demo, peeking at the "claude agents documentation" session shows: "16m explained Agent View -- parallel session manager in Claude Code CLI, opened via left-arrow or claude agents, supports peek/inline-reply/background launch." From here you can reply directly, open fully, or delete the session with Ctrl+X.

Peek view showing a session summary with reply box, and controls for opening, closing, and deleting

Bringing in External Sessions with /bg

If you already have a Claude Code session running in another terminal window, you can bring it into Agent View without starting over. Start Claude in a separate window, give it a task, then type /bg (background). The session disappears from that window and appears in the Agent View dashboard. In the demo, a "brainstorm new Claude agents title ideas" session shows up within seconds of being backgrounded.

A separate terminal window with Claude Code open, about to type a new prompt to background into the agents dashboard

Starting New Sessions From the Dashboard

You can also launch sessions directly from Agent View. A prompt at the bottom reads "describe a task for a new session." Type a task like "create a landing page talking about Claude agents" and it immediately appears under Working. In theory, you could run your entire workflow from this single view and never leave it.

Agent View showing a newly created "create a landing" session alongside other working and completed tasks

Persistence and Safety

Exiting the Agent View window does not kill your sessions. Close it, reopen it with claude agents, and the same dashboard reappears with all sessions intact. This means you are not accidentally shutting off eight sessions at once if you close the window -- a critical detail for anyone running long-running tasks.

Key Takeaways

Agent View (claude agents) is a unified dashboard for managing multiple concurrent Claude Code sessions, organized by status: needs input, working, and completed.
Peek and reply (spacebar) lets you check on and respond to any session without fully switching into it, reducing context-switch overhead.
Background mode (/bg) moves any existing Claude Code session into Agent View, and new sessions can be started directly from the dashboard.
Sessions persist across Agent View restarts -- closing the window does not terminate running tasks.

Resources

Claude Code CLI -- Anthropic's official CLI tool for Claude
@chase_ai_ on TikTok -- Chase AI's channel covering AI developer tools