Anthropic has revamped the Claude desktop app. The update adds interface changes, new session management tools, design and document features, and customization options. A lead engineer at Anthropic described the release as Claude 2.0. This article summarizes the eleven most significant changes.

Structural Changes

The desktop app now shows only two tabs, Chat and Code. Co-work has been folded into Chat as of the previous month. Anthropic positions Claude Code as its primary product and ships most updates there. Claude Code does not require programming skills. Users direct Claude in plain English.

1. The Session Rail

The session rail is a panel on the right side of the window. If it is hidden, users can select Show session rail in the top-right corner of the interface. It contains a set of cards that describe the current session:

• Environment card: Shows the folder in which the session works and whether it runs on the user's computer or in the cloud. A remote control toggle allows users to continue the same session from a phone.
• Context card: Shows how much of the context window is in use and what occupies it. The newest Claude models have a 1-million-token context window, roughly equal to four Harry Potter books placed side by side. Apart from user prompts, the card displays memory files, skills, and MCP connectors. These elements are preloaded into every new chat and count against usage limits from the first prompt. The card helps users identify files that have grown too large and need cleanup.
• Connectors card: Lists the apps connected to Claude.
• Outputs card: Lists every file the session has created.

Settings adjustments:

• Under Settings > Claude Code, enabling connect new sessions to remote control makes every new session available on the phone without manual activation.
• Enabling keep computer awake while Claude works prevents the computer from sleeping during a task.

2. Sidebar Sections

Chats can be grouped into sections that function like folders. To create one, users select the three-dot menu next to a chat, choose Move to new section, and name the section. Sessions can then be dragged into the section. Users can organize chats by project, client, or business function.

3. Nested Sessions

A session can start its own sub-sessions. These appear beneath the parent session in the sidebar. When a sub-session finishes, it reports back to the session that started it. For example, a marketing campaign could use one main session that starts three sub-sessions: one for a sales page, one for emails, and one for ads.

Nested sessions differ from sub-agents. A sub-agent receives a prompt and returns a result. Users can view its progress but cannot interact with it directly. Each nested session is a full chat in the sidebar. Users can open it, review every action it performed, and continue working in it directly.

4. Master Sessions

In Claude Code, one session can view the user's other sessions. A user can designate one session as the master session and pin it to the top of the sidebar. From the master session, a user can:

• Ask which decisions other chats are waiting on. Claude reviews those chats, identifies those that require user input, and summarizes them.
• Start new sessions and control them from the master session.
• Ask Claude to group existing chats into sections. In the example shown, Claude created a section for all chats related to a specific video without manual steps.

By default, a master session can see the 20 most recently active sessions. Users can instruct Claude to search further back.

5. Claude Code Projects

Projects are similar to nested sessions, with three key differences:

• Execution location: Projects run in the cloud and continue working when the user's device is closed. Nested sessions run on the user's computer and stop when the computer is asleep.
• Shared memory: A rule set once applies to every chat session within the project.
• Overview page: The page shows completed work, active work, and items awaiting user input.

Access to files and skills: Nested sessions can access the user's local skills, files, and tools. Projects can use them only after the user adds them to the project. The recommended method is to ask Claude to collect the required skills and context from the computer while it remains awake. Projects run on Anthropic's cloud, so any business file uploaded to a project is stored on Anthropic's servers.

Selection guidance: Use nested sessions for work that requires the local computer or local files. Use projects for large, ongoing work that must continue while the user is away.

Availability: Nested sessions are available on all plans. Projects are in beta and available only on the Pro and Max plans. If Projects does not appear in the sidebar, users should check under More. If it is not available there, users may need to join the waitlist.

6. Side Chats

When a question arises during a long task, users can open a side chat without interrupting the main session. To open one, select the three-dot menu at the top right or type /btw. The side chat opens next to the main session and can read the entire main session. Users can ask it to explain completed work or check task status.

Previously, side chats were temporary and were not saved. Users can now open multiple side chats, and each one is saved for later access.

7. Split View

Split view displays several chats on one screen. To use it:

• Hold Alt (Windows) or Option (Mac) and select a chat in the sidebar, or
• Right-click a chat and select Open in split view.

Users can zoom out with Ctrl + minus (Windows) or Command + minus (Mac). The presenter displayed up to 10 chats on one screen. Many simultaneous panes increase the cognitive load on the user. Beginners should start with two chats and add more as they become comfortable.

8. Artifacts: Docs, Slides, Design, and Dashboards

Claude now includes tools that create artifacts. Each artifact is a shared workspace in which the user and Claude work on the same file. Users can leave comments for Claude to act on. Artifacts can be published as live links or shared with teams and clients.

Design system support: Users can upload their own design systems in Claude Code settings. Alternatively, they can use a skill that imports the design system.

Plan availability: Pro and Max subscribers have access to all artifact types except Motion, which is not yet available on those plans. Anthropic is expected to add it.

9. AI-Powered Artifacts

AI-powered artifacts can include Claude functionality inside them. To enable this, users open Settings > Capabilities and switch on AI-powered artifacts.

Users can ask Claude to add a chat window or search bar to a shared dashboard. Team members can then type questions and receive answers based on the dashboard data. Team members do not need to use Claude Code to query the data. They open the shared link and ask questions directly within the artifact.

10. Slash Commands

A slash command is a shortcut typed into the chat. Anthropic has added new commands and updated several existing ones.

11. Mods

A mod is a small add-on that changes how the Claude Code app looks or works. Users create mods by describing the desired change to Claude. The presenter highlighted two examples:

• Ribbon: A slim bar above the chat input. It displays remaining weekly usage and shows whether the prompt cache for the current chat is warm or cold. This is free and open source.
• Savvy Progress: Adds a progress bar above the chat input, showing how close Claude is to completing a long task. For tasks that use sub-agents, it also displays a live panel for each agent.

Prompt cache: The prompt cache is a short-term memory that lets Claude avoid rereading the full conversation with each message. After approximately one hour, the cache expires. The next message then forces Claude to reread the entire conversation at standard cost, which can consume usage limits quickly. Monitoring cache status can reduce token use over time.

Keyboard Shortcuts