Anthropic Ships Claude Code Agent View for Parallel Multi-Session Orchestration
Anthropic released "Agent View" for Claude Code on May 12 as a research preview — a unified terminal dashboard that consolidates all parallel Claude Code sessions into a single pane. Four independent sources confirmed the same-day launch, with unusually broad practitioner coverage indicating immediate uptake among power users.
What the Source Actually Says
Agent View is accessed by pressing the left arrow inside any active Claude Code session, or by running claude agents in a fresh terminal. The resulting dashboard lists every running session with live status indicators: yellow signals a session waiting for human input, green marks completion, and no color means the session is still running. Developers navigate with arrow keys, reply inline to unblock waiting sessions without losing their place, and background the current session at any time with /bg. The claude --bg "<task>" flag launches detached sessions scoped to specific working directories, enabling cross-project orchestration from one view.
Nate Herk's same-day walkthrough captures the practical delta precisely: the previous workflow for running four to ten parallel Claude Code sessions meant managing VS Code terminal tabs with no visibility into agent state. Hooks existed as a notification workaround, but the tab-switching overhead remained. Agent View removes that entirely. Additional capabilities ship alongside: /goal "<objective>" starts a long-running Karpathy-style optimization loop — sessions run overnight against an objective metric and surface their status in the same Agent View. Ctrl-X twice kills any session. AlphaSignal notes the feature requires Claude Code v2.1.139+ and that parallel sessions each count against the rate limit equally.
Anthropic's official @claudeai account confirmed availability on all paid plans the same hour. Boris Cherny, Claude Code's engineering lead, was cited in a concurrent Nate Herk deep-dive as the canonical Level 4 workflow example — five parallel sessions across numbered terminal tabs. Agent View now makes that the default experience rather than an advanced workaround.
Strategic Take
Agent View repositions Claude Code as a fleet manager rather than a single-session coding assistant, reducing the cognitive overhead of parallel agentic work to near-zero input. For teams evaluating Claude Code for multi-task pipelines, the remaining friction is the research-preview stability caveat — performance regressions were noted on launch day — but the UX gap with purpose-built orchestration dashboards has materially narrowed.
