Claude Code desktop update

Anthropic has redesigned Claude Code’s desktop app around parallel work, adding side-by-side sessions and a new sidebar to manage them in one window. (code.claude.com) The updated desktop interface adds a drag-and-drop pane layout, an integrated terminal, an in-app file editor, visual diff review, and live previews for development servers, HyperText Markup Language files, and Portable Document Format files. Anthropic says those layout and editing features require Claude Desktop version 1.2581.0 or later. (code.claude.com) Claude Code’s desktop app now runs sessions in separate Git worktrees, which lets users work on multiple tasks in parallel without mixing file changes across branches. The desktop quickstart also says the app supports local sessions, Secure Shell connections, and Anthropic-hosted remote sessions that can continue after the app closes. (code.claude.com, code.claude.com) Anthropic paired the desktop redesign with “routines,” a research-preview feature that packages a prompt, repositories, and connectors into a saved job that runs automatically. Anthropic says routines can trigger on a schedule, from an Hypertext Transfer Protocol application programming interface call, or from GitHub events such as pushes, pull requests, issues, and workflow runs. (code.claude.com) Those routines run on Anthropic-managed cloud infrastructure rather than the user’s laptop, so they keep working when the computer is closed. Anthropic says routines are available on Pro, Max, Team, and Enterprise plans with Claude Code on the web enabled. (code.claude.com) Anthropic has also been pushing Claude Code onto more surfaces than the terminal. Its product page now positions desktop, browser, mobile, integrated development environments, and Slack as parts of one system, with web and iOS tasks in research preview and mobile-to-desktop handoff available on Pro and Max plans. (code.claude.com, code.claude.com) The underlying pitch is that coding agents are moving from chat windows to longer-running workspaces. Anthropic’s documentation describes Claude Code as an “agentic” tool that reads a codebase, edits files, runs commands, and verifies results, while the desktop app adds visual controls for people who do not want to stay in a terminal all day. (code.claude.com, code.claude.com) Microsoft is making a similar case for GitHub Copilot Command Line Interface, but from inside the shell. GitHub’s product page says Copilot CLI can move from “/plan” to implementation, use “autopilot” for less step-by-step approval, and dispatch parallel subagents with “/fleet.” (github.com, docs.github.com) GitHub’s documentation says “/fleet” breaks a larger request into smaller tasks and runs them in parallel where dependencies allow. Microsoft highlighted that workflow this week in a demo that showed Copilot CLI building a cricket-scoring app in under 30 seconds using “/plan,” “autopilot,” and “/fleet.” (docs.github.com, github.com) Anthropic and GitHub are now competing on the same promise: one agent can draft the plan, several subagents can do the work, and the human reviews the result in a desktop app, browser, or terminal. The new Claude Code release pushes that model further toward always-on software work instead of one prompt at a time. (code.claude.com, docs.github.com)