Anthropic dropped a major redesign of the Claude Code desktop app today, and it's not just a visual refresh. The whole thing has been rethought around one idea: you're not waiting on one task anymore — you're orchestrating many.
If you've been using Claude Code seriously, you know the friction. Multiple repos, multiple tasks, alt-tabbing between your terminal and the app, losing context mid-session. This update addresses most of that directly.
Here's what actually changed.
1. Sidebar With Multi-Project Session Management
The new sidebar groups all your sessions by project. Every repo you've worked on shows up with its sessions listed underneath — you can jump between them instantly without losing context.
This is the biggest UX shift. Before, managing multiple parallel workstreams felt clunky. Now it's front and center. Kick off a refactor in one repo, a bug fix in another, and switch between them as results come in.
2. Integrated Terminal — No More Alt-Tabbing
The terminal is now a built-in pane inside the app itself. You can have your chat session on the left and a live terminal on the right, docked together in one window.
This sounds small, but it's significant in practice. Before, you kept a separate terminal open on the side. Now everything lives in one place — run commands, check output, continue the conversation, all without switching windows.
3. Side Chat — Ask Without Interrupting the Agent
This is my personal favorite addition. Press Ctrl + ; (or ⌘ + ; on Mac) during any active session to open a side chat.
The key detail — straight from the UI itself:
Claude sees the full session context, and nothing here is added to the main conversation.
So you can ask "wait, what does this function actually do?" or "is this approach correct?" without derailing the agent mid-task. It's context-aware but isolated. Genuinely useful.
4. Model + Effort Control in One Place
Click the model selector at the bottom right, and you get a clean dropdown with two controls together: model and effort level.
The Effort setting (Low / Medium / High) is underrated. Medium is the default sweet spot — fast enough, thorough enough. Switch to High when you need Claude to really think through a complex architectural decision. Switch to Low for quick edits and boilerplate.
Having model + effort together in one click is clean. No digging through settings.
Other Notable Changes
Beyond the four big ones:
- Drag-and-drop pane layout — arrange terminal, preview, diff viewer, and chat in any grid you want
- Faster diff viewer — rebuilt for large changesets, noticeably snappier
- Three view modes — Verbose, Normal, Summary — dial from full transparency into Claude's tool calls down to just the results
- Plugin parity with CLI — desktop app now fully supports Claude Code plugins
- SSH support on Mac — previously Linux only
- Expanded preview pane — open HTML files and PDFs directly in-app
The Bigger Picture
The old Claude Code desktop felt like a wrapper around the CLI. This redesign feels like a proper IDE for agentic work — one where you're managing agents, not just chatting with one.
The parallel session management, integrated terminal, and side chat together tell a coherent story: you're the orchestrator now, and the UI should reflect that.
If you're on Pro, Max, Team, or Enterprise — update and restart. It's worth it.
Using Claude Code heavily? Drop your workflow in the comments — curious how others are structuring multi-session work.
Top comments (0)