What is Agent View and How It Works

Agent View is a new feature in Claude Code that gives developers a central overview of all running agent sessions. Instead of switching between terminal windows, you see all your Claude Code agents in a single table, their status, and their latest responses. It works directly in the CLI — no GUI, no browser, just a pure command line.

You activate Agent View in two ways: either by pressing the left arrow key from any running session, or with the command claude agents from the terminal. Each row in the overview shows the session name, whether the agent needs something from you, the latest response, and the time of the last interaction.

Three Key Features

1. Overview of everything in one place. Agent View displays all parallel agents in a simple table. You immediately see who is waiting for your input, who is working, and who is done. No wandering between terminal tabs.

2. Preview and respond without leaving your current session. By hovering over any session, you can see the latest interaction. If an agent is waiting for a decision, you can respond directly in the preview — the session continues without you leaving what you are currently working on. Press Enter to connect directly to the session where you want to explore the full transcript.

3. Sending tasks to the background. You can move an existing session to the background with the command /bg. You can launch a new agent directly in the background using claude --bg [task]. The agent works while you do something else, and you return to it when it is done or needs your attention.

How Developers Use Agent View in Practice

Based on feedback from early users, Anthropic identified several typical workflows:

Parallel idea execution. Developers send several different approaches at once — each agent gets a different solution variant, or a different skill. In a few minutes, they return to a list of pull requests ready for review. Instead of hours of experimenting, you get several complete solutions in a few minutes.

Managing long-running agents. Agents that monitor pull requests, update dashboards, or perform periodic code checks run in Agent View with the time of the next scheduled run displayed. You don't have to hide them anywhere — they simply run in the background.

Quick context switching. A main task in progress, but you need to quickly find out something about the codebase? Left arrow, you launch a side query, right arrow — you're back. When the side agent responds, you see the result in the preview.

Overview of completed work. Status indicators on each row plus the task name in the preview allow for instant checking of which agents created a pull request and what exactly they delivered.

Availability and Limits

Agent View is available from May 11, 2026 as a Research Preview on all paid Claude plans: Pro ($20 per month, approx. 460 CZK), Max (from $100 per month), Team ($25 per user per month), and Enterprise. The feature is also available for Claude API users.

For Czech developers and companies, it is important that Claude Code is fully available in the Czech Republic — Anthropic supports the Czech Republic as one of the countries where Claude services can be used without restrictions. Prices are listed in dollars, but payment is made by card in the standard way without regional surcharges.

Activation requires only running the claude agents command in the terminal. Standard rate limits apply — Agent View does not increase the allowed number of requests, it only simplifies the management of parallel sessions. Detailed documentation is available at code.claude.com/docs/en/agent-view.

Why It Matters

Agent View is not a cosmetic change — it reflects a broader shift in software development. As agent AI becomes a common part of developer practice, the need for tools that enable orchestrating multiple agents simultaneously grows. A single AI agent is useful, but a team of parallel agents managed from one place changes the way developers think about work.

With this step, Anthropic responds to the reality that Claude Code developers live: one agent is not enough. When you need to solve refactoring, write tests, check security, and generate documentation in parallel, you need an overview — and that is exactly what Agent View offers.

For Czech development teams that are experimenting with Claude Code or have already deployed it in their work, Agent View represents a practical productivity boost without having to learn new tools. Everything stays in the terminal, exactly where developers spend most of their time.