Move Your Session Anywhere with /teleport

Automate Repetitive Tasks with /loop and /schedule

Hooks for Deterministic Automation

Git Worktrees for Parallel Sessions

Voice Input with /voice

The Chrome Extension for Frontend Work

/branch and --fork-session for Experiments

/btw for Side Questions

--bare for SDK Speed

--add-dir for Multi-Repo Work

Custom Agents with --agent

What This Means for DevOps

Try It Out Most people use Claude Code to write code, fix bugs, and maybe generate a commit message. That's fine, but you're leaving a lot on the table. Boris Cherny, the creator of Claude Code, recently shared a thread on X about features that even daily users tend to overlook. Some of these genuinely changed how I work. Here's a rundown of the ones worth knowing about. Claude Code has mobile sessions, automated scheduling, voice input, parallel agents, git worktrees, hooks, and a browser extension. Most people use about 20% of what it can do. You can start a session on your laptop and pick it up on your phone. Or move it to the web. The /teleport command transfers your full session context between devices. The reverse also works. If you're reviewing something on your phone during a commute, you can /teleport it back to your terminal when you sit down. There's also /remote-control which lets you connect to a running session from another device without transferring it. This is useful when you kick off a long-running task on your workstation and want to check progress from your phone. This one is a genuine workflow changer. You can tell Claude Code to run a task on a recurring schedule for up to a week. Think about what you do repeatedly: reviewing PRs, checking CI status, monitoring deployments, updating dependencies. You can automate all of it without writing a single script. Some practical examples: Hooks let you run code at specific points in Claude Code's lifecycle. Unlike the AI-driven /loop command, hooks are deterministic - they always run the same way. You configure them in your settings and they fire on events like: This is powerful for teams. You can enforce standards (like running linters before every commit) without relying on each engineer to remember. If you've ever wanted Claude to work on two different branches at the same time, worktrees make this possible. Each session gets its own isolated copy of the repo. Why this matters: you can have Claude refactoring module A while simultaneously building feature B. Neither session interferes with the other. This pairs well with /batch, which fans out work across dozens of parallel agents. Need to update 50 files? /batch can process them concurrently instead of one at a time. You can dictate to Claude instead of typing. This sounds gimmicky until you try it for longer explanations. It's particularly useful for: Typing detailed prompts takes time. Talking is faster for anything longer than a few sentences. Claude Code has a Chrome extension that lets the AI see what your app looks like in the browser. Instead of describing UI bugs, Claude can verify its own output visually. This closes the feedback loop for frontend work. Claude makes a change, checks the browser, adjusts if something looks off. You stop being the human screenshot tool. Want to try two different approaches to the same problem? /branch creates a copy of your current session so you can explore a different path without losing your progress. This is like git branches but for your AI conversation. Try approach A in one branch, approach B in another, then pick the winner. When Claude is working on a long task, you might have an unrelated question. Instead of interrupting the main task, /btw lets you ask a side question. Claude answers your side question and goes right back to what it was doing. No context switching, no lost progress. If you're using Claude Code in scripts or CI pipelines, the --bare flag skips loading plugins and extra features, making startup up to 10x faster. This matters when you're calling Claude from automation scripts where every second counts. Working across multiple repositories? You can give Claude access to all of them in a single session. Now Claude can see your API schema and your frontend code at the same time. No more copying types between repos or explaining your API structure manually. You can create custom agent configurations with their own system prompts and tool permissions. Define these in your .claude/agents/ directory. Each agent can have different instructions, different tool access, and different behaviors. A code reviewer agent doesn't need write access. A deployment agent doesn't need to browse the web. These features shift Claude Code from "AI code assistant" to "AI DevOps team member." The combination of scheduling, hooks, parallel sessions, and multi-repo access means you can automate workflows that previously required custom tooling. Here's a realistic DevOps setup: The key insight from Boris's thread: "There is no one right way to use Claude Code." The tool is intentionally flexible. Experiment with these features and build the workflow that fits your team. If you haven't updated Claude Code recently, run: Many of these features are recent additions. The mobile app, scheduling, and hooks in particular have been added in the last few months. For more DevOps tools and guides, check out our exercises and quizzes to sharpen your skills. This post was inspired by Boris Cherny's thread on X. Boris is the creator of Claude Code at Anthropic. Templates let you quickly answer FAQs or store snippets for re-use. Are you sure you want to ? It will become hidden in your post, but will still be visible via the comment's permalink. Hide child comments as well For further actions, you may consider blocking this person and/or reporting abuse