Claude Code adds /rewind command and fixes scroll behavior

By Harsh Desai4 July 2026

TL;DR

Claude Code adds support for the /rewind command to resume conversations from before a clear. The update fixes scroll position jumping during streaming responses and prevents stopped background agents from resurrecting.

What changed

Claude code now includes a /rewind command that lets developers and vibe builders resume conversations exactly from before any clear action. The update also stops scroll position from jumping during streaming responses for basic users and prevents stopped background agents from coming back unexpectedly.

Why it matters

Vibe builders gain reliable ways to pick up extended sessions without starting over while developers avoid losing thread context in iterative work. In use cases like multi turn debugging flows this matches the reliability seen in tools such as Cursor where session continuity keeps 80 percent of prior steps intact.

What to watch for

Compare the behavior against OpenAI Codex when testing long chats. Run a clear command followed by /rewind then stream a response to verify the scroll stays stable throughout.

Who this matters for

Vibe Builders: Use the /rewind command to recover context after a clear and maintain long creative sessions.

Harsh’s take

The addition of /rewind addresses a major friction point in terminal-based AI workflows: the accidental loss of context. For operators who rely on high-velocity prompting, the ability to undo a 'clear' action brings Claude Code closer to the stateful reliability of IDE-integrated tools like Cursor. Fixing the scroll-jump bug is more than a UI polish.

It prevents the cognitive load of chasing text while it streams, which is essential for reviewing code changes in real-time. These stability updates signal that Anthropic is prioritizing the developer experience for long-running, multi-turn sessions rather than just short, disposable queries.

by Harsh Desai

Source:code.claude.com