Anthropic just gave its Claude Code desktop app the kind of overhaul that makes it feel less like “an AI in a window” and more like a proper control room for agentic coding. Instead of nudging you toward one prompt at a time, the new design leans into the reality of how people are actually using AI for code now: lots of parallel tasks, multiple repos, and you in the middle trying to keep everything on track.

Open the app now and the first thing you’ll notice is the new sidebar. Every active and recent session shows up there, so you can kick off a refactor in one project, a test-writing pass in another, and a bug fix somewhere else, then bounce between them as results land. You can filter sessions by status, project, or environment, or group them by project when you’re deep inside a single codebase. When a session’s pull request gets merged or closed, it automatically archives itself, which keeps the sidebar focused on what’s actually live instead of becoming a graveyard of old experiments.

Anthropic is clearly designing around “parallel agents” as the norm, not the edge case. The company describes today’s agentic workflow as having “many things in flight, and you in the orchestrator seat,” and the app now mirrors that mental model. You’re not meant to fire off one big request and wait; you’re meant to spin up multiple workstreams, watch them move, and intervene when something drifts.

To support that, the desktop app pulls more of your workflow inside instead of constantly punting you back to your editor or terminal. There’s an integrated terminal so you can run tests or builds alongside the session that proposed the changes. A built‑in file editor lets you open files, make quick edits, and save directly, which is especially handy when Claude’s suggestion is 90 percent right but not quite shippable. The diff viewer has been rebuilt to handle larger changesets more smoothly, and an expanded preview pane can now render HTML and PDFs in the app, on top of the existing ability to run local app servers there. All of this lives in a drag‑and‑drop layout, so you can organize chat, diffs, terminal, and preview panes into whatever grid feels natural for the way you work.

One subtle but important addition is side chat. With a simple keyboard shortcut (Command+; on macOS or Ctrl+; on Windows), you can branch off a quick conversation mid‑task without polluting the main thread that’s steering your agent. Side chats inherit context from the primary conversation but don’t push anything back into it, which is a smart guardrail: you can ask “what does this helper actually do?” or “could we try an alternative approach?” without accidentally changing the instructions that are driving your ongoing work.

On the platform side, Anthropic is trying to make Claude Code feel like a first‑class citizen alongside the CLI. The desktop app now has parity with CLI plugins, so whether your organization manages plugins centrally or you’ve installed them locally, they’ll behave the same way in the app as they do in your terminal. Sessions can still run locally or in the cloud, and SSH support has been extended to Mac as well as Linux, which means you can point Claude Code at remote machines from either platform. For teams that have already wired Claude Code into their stack through plugins and remote environments, this reduces the mental friction of switching into the desktop UI.

The interface itself is more tunable, too. There are three view modes—Verbose, Normal, and Summary—so you can decide how much of Claude’s internal tool‑calling behavior you actually want to see. If you like watching each step of the reasoning and tool usage, Verbose mode is there; if you just want the final results, Summary mode hides most of the plumbing. New keyboard shortcuts cover session spawning, switching, and navigation, and a quick Command+/ or Ctrl+/ brings up the full list so you don’t have to memorize everything on day one. There’s also a new usage button that surfaces both your context window and session usage at a glance, which should help power users avoid accidentally bumping into limits mid-flow.

Under the hood, Anthropic says the app has been rebuilt for reliability and speed, and now streams responses as Claude generates them instead of making you wait for full blocks of text. That streaming behavior, combined with the parallel session design, should make the whole experience feel more responsive, especially when you’ve got multiple agents chewing on different parts of a large codebase.

Availability wise, this isn’t a niche beta. The redesigned desktop app is rolling out to all Claude Code users on Pro, Max, Team, and Enterprise plans, and it’s also accessible via the Claude API. If you already have the app installed, it’s just a matter of updating and restarting; new users can grab it from Anthropic’s download page and then dive into the Claude Code docs to understand everything from session management to the new routines features that launched alongside it.

The bigger story here is how quickly AI coding tools are moving past the “single chat box” paradigm. Anthropic is betting that serious developers want a cockpit where multiple agents can work in parallel, plugged into real infrastructure, with enough transparency and control to ship code confidently. This redesign of the Claude Code desktop app is a step in that direction: less of an assistant on the side, more of a programmable co-worker that fits into the way modern teams already build software.