Boris Cherny, Core Developer of Claude Code, Shares 15 Often-Overlooked Advanced Features
Recently, Boris Cherny, considered a key figure in the Claude Code project, compiled and shared a list of 15 advanced features. While incredibly useful, these features are often overlooked by the broader user base. The list covers multiple dimensions including cross-platform development, task automation, workflow optimization, and environment customization, providing developers with a guide to more effectively utilize this AI coding assistant.
Seamless Cross-Platform Development and Collaboration
To accommodate the multi-device, multi-environment work model of modern developers, Claude Code offers a series of features to ensure development continuity and flexibility.
- Mobile App Support: Claude Code has released official iOS and Android apps, allowing developers to review and modify code directly on their mobile devices, enabling true mobile development.
- Seamless Session Migration: By running the
/teleport or claude --teleport command in the terminal, developers can seamlessly transfer a cloud session to their local machine to continue working. Conversely, the /remote-control command allows users to remotely control a locally running session via a mobile phone or web page.
- Parallel Session Forking: To support exploring multiple solutions for the same task, users can “fork” a separate, parallel session from the current one using the
/branch command or claude --resume <session-id> --fork-session instruction, without affecting the main process.
- Non-Interruptive Queries: While an Agent is executing a task, developers can use the
/btw (By The Way) command to interject questions, retrieve information, or perform quick lookups without interrupting the Agent’s primary ongoing task.
Deep Automation and Task Scheduling
Automation is key to boosting development efficiency. Claude Code has powerful built-in automation tools that can delegate repetitive work to the AI.
- Recurring Task Scheduling:
/loop and /schedule are two core automation commands that allow users to set up recurring tasks for up to a week. For example, using /loop 5m /babysit can make a Claude Agent automatically perform a preset task like a code review or rebase every 5 minutes.
- Lifecycle Hooks: Developers can inject custom logic at different stages of an Agent’s lifecycle through Hooks. For instance, use the
SessionStart hook to dynamically load context at each startup; use PreToolUse to log every bash command executed by the model; or use PermissionRequest to forward permission requests to a specified application for approval.
- Remote Desktop Control (Cowork Dispatch): As a security feature of the Claude Desktop App, Dispatch allows an authorized Agent to remotely control the user’s computer to handle Slack messages, manage emails and files, enabling responses to important matters even when the user is away from their computer.
- Large-Scale Batch Processing: For parallelizable large-scale tasks like code migrations, the
/batch command can automatically distribute the work to hundreds or thousands of parallel worktree agents, significantly improving processing efficiency.
Workflow Optimization and Environment Customization
To meet the specific needs of different projects and teams, Claude Code provides a wealth of customization options, allowing developers to fine-tune their work environment.
- Deep Git Worktrees Integration: Git Worktrees allow for handling multiple tasks in parallel within the same repository. Claude Code offers deep support for this, which users can enable via the
claude -w command or by checking the corresponding option in the desktop app. This is crucial for complex workflows managing dozens of Claude instances simultaneously.
- Frontend Development and Testing: The official recommendation is to use its Chrome extension for frontend development, as it provides a reliable environment for Claude to validate and iterate. Meanwhile, the Desktop App has built-in capabilities to start and test web services, allowing it to preview and debug within an integrated browser, creating a closed-loop development cycle.
- Performance Optimization: For non-interactive scripts, using the
--bare startup parameter can skip the search for local configuration files and the MCP, boosting SDK startup speed by up to 10x. Future versions plan to make this the default setting.
- Extended Access Permissions: When working on cross-repository projects, you can use the
--add-dir command or type /add-dir in a session to temporarily grant a Claude Agent permission to access and operate on other directories. This configuration can also be made permanent by adding additionalDirectories to the settings.json file.
- Custom Agents: By defining new Agent configuration files (including custom system prompts and tools) in the
.claude/agents directory, developers can use the claude --agent=<agent-name> command to invoke custom Agents with specific capabilities, meeting highly customized needs.
Multimodal Interaction
To further lower the barrier of human-computer interaction, Claude Code has also introduced voice input functionality.
- Voice Input: Boris Cherny himself states that he completes most of his coding work through voice. Users can run the
/voice command in the CLI and then hold the spacebar for voice input, or directly use the voice button in the desktop app and the dictation feature on iOS, achieving a “hands-free” programming experience.
