Claude Code has quietly accumulated a set of commands that go well beyond basic chat-and-code. Most developers stick to the defaults, but a few of these are worth adding to your muscle memory.
The Ones That Save Real Time
/btw lets you ask Claude a question mid-task without derailing what it's already working on. If Claude is halfway through a refactor and you need to check something unrelated, /btw runs a parallel query that gets cleared separately from the main conversation. No context pollution, and it reuses the prompt cache so you're not burning extra tokens.
/model opusplan is available to $20 Pro subscribers and does something clever: it routes complex reasoning to Claude Opus 4.6, then drops back to Sonnet 4.6 for execution. You get Opus-quality planning without blowing through your quota on straightforward code generation.
/rewind (or just hit Escape twice) now supports granular rollback. You can revert just the code, just the conversation, or both. You can also compress the conversation from a specific point to reclaim context window space - useful when you're deep into a long session and Claude starts losing track of earlier decisions.
Workflow Commands
/branch duplicates your current conversation into an independent session. Want to try two different implementation approaches without losing your place? Branch it. This was previously called /fork.
/loop repeats a task on a timer, defaulting to every 10 minutes. Tasks auto-expire after three days. Practical use: /loop 5m check deployment status to babysit a deploy without switching windows.
/simplify replaced the older /review command in late February. It launches three parallel agents that check your changed code for reusability, quality, and efficiency all at once.
/insights generates an HTML report of your Claude Code usage over the past month - which commands you lean on, where you're repeating yourself, and what custom commands or skills might help. Good for a monthly self-audit.
A couple of keyboard shortcuts are easy to miss too: Ctrl+V pastes screenshots directly into the terminal (yes, even on Mac - it's Ctrl, not Cmd), and Ctrl+R searches your prompt history the same way it works in a regular shell.
None of these are new releases. They've been shipping incrementally over the past few months. But taken together, they represent a meaningful gap between "using Claude Code" and "using Claude Code well."