Free Your Conversation: /code-review Now Runs in the Background in Claude Code

Since v2.1.218 (July 22), /code-review launches as a background subagent — reviews no longer fill your conversation thread, and you can keep coding or stack commands while one runs.

What changed

Before v2.1.218, running /code-review consumed your conversation context and blocked the thread until the review completed. Now it runs in a separate background subagent:

• The conversation stays clean — no wall of review output mid-session
• You can continue coding, file edits, or other commands while the review is in progress
• Stacked slash commands (e.g., /code-review ultra my-branch) pass the trailing text as a targeting note to the background review rather than triggering a second skill

Basic usage

# Review your current diff (background by default)
/code-review

# Pass a focus note — Claude factors it into the review findings
/code-review focus on security in the auth module

# Review a specific branch or PR number
/code-review feature/api-refactor
/code-review 214

The background review agent reads your staged/unstaged diff, cross-checks it against your CLAUDE.md conventions, and writes its findings to the session transcript. A notification appears in the status bar when it finishes.

When it still runs in the foreground

Two cases stay synchronous:

1. Non-interactive mode (claude -p / Agent SDK) — background subagents don't apply here
2. A second call while the first is still writing — it waits for the current slot to free up

Pair it with /ultrareview for deeper coverage

/ultrareview spawns multiple parallel review subagents (correctness, security, performance dimensions) and is still a separate command. Use /code-review for a fast background pass; reach for /ultrareview when you want adversarial coverage across dimensions on a large diff.

Sources: Code Review — Claude Code Docs · Claude Code v2.1.218 Release Notes (GitHub) · Claude Code 2.1.218: What Shipped (joinnextdev.com) · Claude Code Changelog (claudeupdates.dev)