Claude Code desktop now shows which sessions need you

When you run five Claude Code agents at once, starting them is easy. The hard part is spotting the one that has stopped and is waiting for you. According to a post from ClaudeDevs, the Claude Code desktop app has been reworked around that problem, and its sidebar now shows which sessions are working and which need you.

At a glance

• When you ask Claude to run tasks in parallel, it spawns a nested session for each one, and each reports its result back into the main session so that thread stays readable.
• Saved environments bundle a few folders, an SSH host or the cloud, with an optional worktree per folder, and /btw sends side questions to their own tab while the main session continues.
• The announcement lists features but no limits: it does not say how many sessions can run at once, how nested sessions affect usage, or how conflicts between parallel edits are handled.

Some background if you haven't been following. Before the desktop redesign, running several Claude Code sessions meant separate terminal windows and no single view across them, and the app kept chat, editor and terminal in separate places. According to DevToolPicks, Anthropic shipped the redesigned app on 2026-04-14 with a multi-session sidebar, drag-and-drop panes, an integrated terminal and file editor, and side chat. On the Claude blog, Anthropic described the app as built for "many things in flight, and you in the orchestrator seat."

Parallel tasks run as nested sessions that report back to the main thread

When you ask Claude to run several tasks in parallel, it starts a nested session for each one. Each nested session does its work separately. When it finishes, it reports its result back into your main session, so the main thread holds outcomes instead of every intermediate step.

Claude also suggests follow-up work while it runs, and one click turns a suggestion into its own session. Side questions go through /btw, which opens the question in its own tab while the main session keeps going. The April redesign already had a version of this: side chat on ⌘ + ; or Ctrl + ;. It pulls context from the main thread but adds nothing back, so a stray question cannot pull the task off course.

Saved environments bundle folders, an SSH host or the cloud

Starting a session is quicker too. The new session screen keeps your recent folders one click away, so you can get back to yesterday's project without browsing for it again.

The bigger addition is environments. You can save a setup as an environment and start sessions in it. That setup might be a few folders, an SSH host or the cloud, and you can choose to get a separate worktree for each folder.

The places themselves were already supported. According to the Claude blog, the April redesign let sessions run locally or in the cloud and extended SSH support to Mac alongside Linux, so sessions could point at remote machines. What is new is saving the combination once instead of rebuilding it for every session. Setup steps are in the desktop quickstart.

The sidebar shows which sessions are working and which need you

The sidebar marks each session as either working or waiting on you. Inside a session, three things track progress. A live to-do list follows Claude's plan as it goes. A recap card sums up the turn when Claude needs you. A side rail shows where the session runs, its context and its changes.

This builds on the earlier redesign. On the Claude blog, Anthropic described the April sidebar as one place for active and recent sessions. You could filter it by status, project or environment, and a session archived itself once its PR merged or closed. The same post introduced Verbose, Normal and Summary view modes and a usage button for context window and session usage.

According to Testing Catalog, this "Claude Hub" experience had been in testing for a while and went through several iterations. The outlet also reports that OpenAI planned a similar experience called "magic notepad" but abandoned it.

How do parallel sessions avoid overwriting each other's files?

They use Git worktrees. A worktree is a second working copy of the same repository, checked out on its own branch and sharing the original's history. According to Build Fast with AI, each session in a Git repository gets its own isolated copy this way, stored by default in <project-root>/.claude/worktrees/. Changes in one session do not affect the others until you commit.

Picture several editors, each with a photocopy of the same manuscript. They can mark up their copy freely, and nothing reaches the master until someone merges it on purpose. Build Fast with AI describes this mechanism as the basis for the per-folder worktree behavior, which environments now offer as an option.

The ClaudeDevs post describes what the hub does but not where it stops. It gives no ceiling on how many sessions can run side by side, and it doesn't say what happens when two nested sessions change the same code and their worktrees have to be merged. In our view, the recap card is the most practical piece, because flagging a session that needs you only helps if you can catch up in seconds. The post doesn't show how that holds up when many sessions stop at once.

What the update instruction leaves open

ClaudeDevs gives one instruction: update the desktop app and start a new session. It names no version number and mentions no staged rollout. The post also doesn't say whether nested sessions, saved environments or the recap card will come to the terminal version of Claude Code. It gives no usage limits for running many sessions in parallel either. Those are the details to look for in the docs and release notes.