Claude Code Feature & SDK Guide FAQs

Question 1

How do I run Claude Code in a headless environment for CI/CD?

Accepted Answer

You can use the --permission-mode bypassPermissions flag along with specific prompt inputs to run Claude Code as a one-shot command or part of a scripted workflow.

Question 2

How can I verify that my snippets are being injected correctly?

Accepted Answer

This skill provides a snippet verification pattern that uses verification hashes and a CLI tool to compare the snippets in your context against the authoritative local list.

Question 3

How do I add external tools to Claude Code?

Accepted Answer

You can add external tools using the Model Context Protocol (MCP) by running 'claude mcp add' followed by the server command, allowing Claude to interact with filesystems, browsers, or APIs.

Question 4

When should I use Claude Opus versus Sonnet in Claude Code?

Accepted Answer

Use Opus for complex reasoning, architectural design, and security audits where deep intelligence is required. Use Sonnet as your default for general coding, debugging, and feature implementation.

Question 5

What is the benefit of using the Claude Agent SDK?

Accepted Answer

The Agent SDK allows you to programmatically build custom agents in Python or TypeScript, enabling you to integrate Claude's reasoning directly into your own applications with custom tools.

Claude Code Feature & SDK Guide

关于

主要功能

使用场景

Claude Code Feature & SDK Guide

关于

主要功能

使用场景