On July 11, Anthropic launched Claude Code Agent View, an integrated CLI console that allows developers to start, monitor, and reply to multiple AI agent sessions within a single interface. Previously, managing five or more concurrent AI agents meant juggling five terminal tabs and a mental task-tracking system — as agent counts grew from one to 20, cognitive load became the real bottleneck, not compute power.
One Screen Replaces Multiple Terminals
Before Agent View, multi-agent workflows in Claude Code relied on developers manually organizing multiple terminal windows, each running a different task, with progress tracked by memory. This approach fell apart once tasks exceeded three or four. The new feature gathers all sessions into a visible list. Developers can press the left arrow key in any session or run claude agents to open the console. Each row displays four pieces of information: session name, whether it's awaiting user input, a summary of the last message, and the time of the last interaction. The "waiting for input" column is critical — previously, if an agent stalled waiting for a decision, developers wouldn't know unless they manually checked each terminal. Agent View surfaces this info prominently, turning multi-agent scheduling into a kanban-like experience.
Background Execution Enables Truly Independent Agents
Another key aspect of Agent View is its background execution mechanism. Developers can push any existing session to the background using the /bg command, or start a new task directly with claude --bg [task description] without entering the foreground interface. The rationale is straightforward: the longer and more complex an AI agent runs, the more developers need a "don't watch it" mechanism. Early user patterns confirm this, with four high-frequency use cases emerging:
First, massive parallel dispatch. Launch multiple tasks simultaneously, each with different skills, then return to the list to browse completed Pull Requests.
Second, long-running task management. For recurring jobs like PR monitoring and dashboard auto-updates, Agent View directly shows the next execution time for each agent.
Third, fast session switching. While in an active task, press left to quickly open a sub-task or code query, then right to return to the main line. The Peek feature shows answers as they arrive without losing focus.
Fourth, tracking deliverables. Status indicators on the list, combined with Peek's title preview, let developers quickly scan which sessions have produced a PR.
All four scenarios point to the same need: as AI agents evolve from tools into semi-autonomous executors, developers shift from writing code to scheduling agents — and scheduling requires a dashboard, not a row of terminals.
Enterprise Adoption Underway
Agent View is currently available as a Research Preview for Pro, Max, Team, Enterprise, and Claude API paid plans, with standard rate limits in effect. Official documentation is live. The number of agents is no longer a problem — the cost of managing them is. Agent View is shifting that cost out of developers' attention budgets.