Unlock 15 Hidden Claude Code Features That Transform Your Development Workflow

Introduction

Many people still view AI coding assistants as merely faster code‑completion tools, but Boris Cherny argues that true productivity gains come from workflow capabilities that keep tasks running continuously and dramatically lower context‑switching costs.

Key Trends Highlighted

Shift from local, single‑machine coding to multi‑device collaboration and remote control.

Shift from manual tool operation to scheduled, automatically looping tasks.

Shift from single‑session serial processing to parallel, large‑scale refactoring with multiple worktrees.

15 Hidden Claude Code Features (Full Breakdown)

1) Mobile Coding

Use the iOS/Android Claude app to edit code without a computer.

Path: download iOS/Android Claude App

Entry: left‑side Code tab

2) Seamless Device Switching

Two commands enable remote session continuation and control: claude --teleport or /teleport – continue a cloud session on a local machine. /remote-control – control a running local session from phone or web.

Reference: https://code.claude.com/docs/en/remote-control

3) /loop and /schedule – Automated Rhythm

These commands let tasks run automatically at defined intervals (up to a week).

Example: /loop 5m /babysit – automatically handle code review and rebase.

Reference: https://code.claude.com/docs/en/scheduled-tasks

4) Hooks – Deterministic Logic in Agent Lifecycle

SessionStart – load context dynamically at start.

PreToolUse – record each bash command.

PermissionRequest – route permission approvals to a messaging channel.

Stop – remind to continue after a stop.

Reference: https://code.claude.com/docs/en/hooks

5) Cowork Dispatch – Remote Takeover

Allows handling Slack, email, file management, and remote notebook control when you are away from the computer.

Dispatch is the secure remote‑control feature of Claude Desktop, capable of invoking MCP, browsers, and local resources after authorization.

Product page: https://claude.com/product/cowork

6) Front‑End Development with Browser Verification Loop

Provide Claude a "verifiable output" environment by attaching a Chrome extension, enabling a write‑view‑adjust‑verify cycle that improves iteration quality.

Reference: https://code.claude.com/docs/en/chrome

7) Desktop Auto‑Launch and Test Web Services

The Desktop app can start services and validate them in a built‑in browser; similar flows are possible via CLI or VSCode with the Chrome extension.

Reference: https://code.claude.com/docs/en/desktop

8) Fork Current Session – Quick Branching

In‑session: /branch CLI:

claude --resume <session-id> --fork-session

9) /btw – Interject Without Breaking Flow

Insert quick side questions while the main task continues; Boris uses this frequently.

10) Deep Git Worktree Usage for Parallel Development

claude -w

– start a new session in a worktree.

Enable worktree option in Desktop UI.

Non‑Git scenarios can use a WorktreeCreate hook for custom logic.

Additional info: https://x.com/bcherny/status/2025007393290272904

11) /batch – Parallel Large‑Scale Refactoring

/batch

clarifies requirements then distributes work across many worktree agents, ideal for massive migrations.

Additional info: https://x.com/bcherny/status/2027534984534544489

12) --bare – Speed Up SDK / Non‑Interactive Scenarios

Disables loading of local configuration; combined with explicit context parameters it can boost startup speed up to tenfold.

13) --add-dir – Cross‑Repository Access

After launching Claude in one repository, use --add-dir (or /add-dir) to extend access to other directories. Teams can set additionalDirectories in settings.json for automatic mounting.

Reference: https://code.claude.com/docs/en/cli-reference

14) --agent – Custom System Prompts and Tool Sets

Package frequently used tasks into a named agent and invoke it with claude --agent=<name>, a powerful way to standardize team workflows.

Reference: https://code.claude.com/docs/en/sub-agents

15) /voice – Voice‑First Coding

Enable voice input across platforms:

CLI: press space after /voice.

Desktop: click the voice button.

iOS: use system dictation.

Product Overview

Claude Code is not a simple "type‑and‑receive" autocomplete; it can decompose a goal into a series of autonomous actions such as reading/writing project files, executing commands, running tests, and handling large‑scale parallel refactoring.

Read/write files and modify across files.

Execute commands, run tests, locate errors.

Combine with scheduled tasks for continuous automation.

Parallel multi‑session handling for massive code transformations.

Two primary usage modes exist: a subscription model for individual developers and an API model for team integration. Pricing starts around $20 / month for the Pro tier, but availability may be limited for users in regions requiring alternative payment methods.

Common Questions

Q1: Which of the 15 features should I try first?

Start with /teleport, /loop, hooks, and worktree —these four have the most immediate impact on task continuity and parallel efficiency.

Q2: Do solo developers need worktree and batch?

Yes. Even single‑person projects benefit from worktree and batch when handling large renames, bulk migrations, or parallel experiment validation, reducing branch‑switching and rollback overhead.

Q3: Why emphasize browser verification loops?

Front‑end work requires visual and interaction correctness, not just successful compilation. Providing Claude a verifiable environment makes iterations more stable.

Q4: When is --bare appropriate?

Ideal for SDK calls or non‑interactive automation scripts where loading local configs would add unnecessary latency; it yields up to a ten‑fold speed increase.

Q5: How can developers in regions with payment restrictions access Claude Code?

Some third‑party services act as proxies to the official API, allowing easier subscription handling, though users should verify compliance with licensing terms.