“Claude 2.0” Is Not an Official Name: What Anthropic's App Overhaul Brings

Anthropic has rebuilt large parts of its Claude desktop app in recent weeks, adding numerous features above all to its programming area, Claude Code. That is according to a German-language YouTube tutorial presenting the changes. The features described are largely backed up by Anthropic's documentation and product announcements, such as the session overview, sub-sessions, Projects and Mods.

The name “Claude 2.0” that has circulated online is therefore not an official one. The video's creator traces the label back to an Anthropic employee who reportedly spoke of “something like version 2”; no independent confirmation could be found.

Cowork folded into the chat

The app now has only two areas. The Claude tab now also contains everything that used to run separately as Cowork. Users would no longer have to choose between chat and Cowork themselves; Claude decides on its own depending on the task. In the Code tab, Claude Code works directly with folders on the computer, reads and writes files and develops programs.

According to the video, new features include a task list in the chat that Claude checks off step by step, a file view that shows changes line by line, and a session bar. Among other things, it shows how full the context window is, which services are connected and which background processes are running.

A “Remote Control” switch lets users continue a session from a smartphone while Claude keeps working on the computer. Claude can control the built-in browser itself. Those who develop iPhone apps will find an iOS simulator in which Claude can test the app.

Sessions launch their own sessions

Claude can now start sub-sessions itself. Unlike the subagents that were already available before, these are said to be full-fledged conversations with their own context, which users can step into at any time. They report their results back to the main session. A session can also read, rename and archive other sessions and send them messages – according to Anthropic, by default it is the 20 most recently active ones.

Running sessions can be moved to Anthropic's servers, where they keep working even when the computer is switched off. The new Claude Code Projects run entirely in the cloud: a coordinator splits a larger goal across threads working in parallel that share a common memory. According to the video, this feature is being rolled out gradually.

New commands and artifacts

The /doctor command checks the Claude Code setup; with the addition “/doctor prompt-audit,” Claude searches instruction files, skills and other configuration for outdated or contradictory instructions. Another command shows how much context each skill takes up. With /insights, Claude creates a report on its own usage, /rewind jumps back to an earlier message, and /recap summarizes the state of a session.

/advisor is intended for users with a tight usage quota. In this mode, a cheaper model such as Sonnet does the work, and Claude brings in a stronger model such as Opus for complex parts. This is meant to use up the quota more slowly.

With so-called Mods, Claude Code can build its own displays, buttons and sidebars on request. Artifacts gained documents, presentations and dashboards that update live, and the design tool Claude Design is now integrated directly into the app. The new Motion area for animated graphics is, according to the video, not yet available to all users.