How many features of Claude Code do you use daily? Coding on your phone? Voice programming? Having Claude automatically handle code reviews every 5 minutes?

Claude Code’s creator, Boris Cherny, listed 15 features he believes are underrated.

Boris says he writes a lot of code on iOS, allowing him to modify code without opening a computer. How to: Download the Claude App (iOS or Android), find the Code tab in the left menu.

The mobile app is better suited for lightweight modifications and task delegation, like starting a task with Claude on the subway, not actually writing complex logic on a 6-inch screen.

2. Sessions Can Switch Between Phone, Web, Desktop, and Terminal

Run claude --teleport in the terminal or type /teleport to pull a cloud session to your local machine and continue. Conversely, use /remote-control to remotely control a locally running session from your phone or web.

Boris has enabled “Enable remote control for all sessions” in the settings.

Docs: Remote Control^[1]

It solves a real pain point: you start a long-running task at your desk, need to leave, but want to check progress and add instructions anytime. Remote Control is still in research preview and can only control one session at a time.

3. /loop and /schedule: Scheduled Automation

Boris says these are Claude Code’s two most powerful features. You can set Claude to run tasks automatically at fixed intervals, lasting up to a week.

He runs several loops locally:

• • /loop 5m /babysit: Automatically handles code review comments, auto-rebase, and shepherds PRs to merge and deployment every 5 minutes.
• • /loop 30m /slack-feedback: Automatically creates PRs based on Slack feedback every 30 minutes.
• • /loop /post-merge-sweeper: Automatically creates follow-up PRs to handle comments missed during previous reviews.
• • /loop 1h /pr-pruner: Cleans up outdated or unnecessary PRs every hour.

Boris’s advice: Turn workflows into a combination of skill + loop.

Docs: Scheduled Tasks^[2]

This might be the most imaginative part of the entire post, turning Claude from a “you ask, it answers” tool into a background-running automated worker. But note some limitations: the terminal must stay open, /loop tasks can run for up to three days (one week is achieved via /schedule on desktop), and the computer cannot sleep. Also, practices like babysit that automatically handle reviews and push to production require good test coverage; it’s not recommended to copy this in projects with weak testing.

4. Hooks: Insert Deterministic Logic into the Agent Lifecycle

Hooks can automatically execute preset logic at specific nodes during Claude’s operation:

• • SessionStart: Dynamically loads context each time it starts.
• • PreToolUse: Logs before each bash command execution.
• • PermissionRequest: Sends permission requests to WhatsApp for remote approval or denial.
• • Stop: Automatically prompts Claude to continue when it stops.

Docs: Hooks^[3]

Hooks are similar to Git’s pre-commit hook but act on the entire lifecycle of the AI agent. This is a key design of Claude Code as an “agent platform” rather than just a “coding assistant.” Advanced users will love it, but the learning curve is not low.

5. Cowork Dispatch: Remote Control of Desktop

Boris says he uses Dispatch daily to handle Slack messages, emails, and manage files, even when not at his computer.

Dispatch is a secure remote control feature of the Claude desktop app that can use your MCP servers, browser, and computer, provided authorization.

Product Page: Cowork Dispatch^[4]</