← 返回命令列表

Linux command

claude 命令

文本

复制后可按需替换文件名、目录或参数。

常用示例

Start an interactive session

claude

Start with an initial prompt

claude "[prompt]"

Run a one-off command

claude -p "[prompt]"

Continue the most recent

claude -c

Resume a specific session

claude -r "[session_id]" "[prompt]"

Pipe input

cat [file.txt] | claude -p "[analyze this code]"

Output response as JSON

claude -p "[prompt]" --output-format json

Specify a different model

claude --model [opus]

说明

Claude Code is Anthropic's official command-line interface for interacting with Claude AI. It provides an agentic coding assistant that can read, write, and execute code directly in your development environment. The tool operates in two main modes: interactive mode for extended sessions with ongoing dialogue, and print mode (-p) for scripted automation and one-off commands. Claude can navigate codebases, edit files, run shell commands, and perform complex multi-step development tasks. Claude Code integrates with your terminal environment, understanding project context through file access and command execution. It maintains conversation history across sessions, allowing you to resume previous discussions. The tool supports piping input from other commands and outputting structured JSON for integration with other tools. Special syntax includes @ for file references and ! for direct shell command execution. The CLI supports MCP (Model Context Protocol) servers for extending capabilities and can connect to IDEs and Chrome for enhanced integration.

参数

-p, --print
Print mode: process prompt and exit without interactive session.
-c, --continue
Continue the most recent conversation.
-r _ID_, --resume _ID_
Resume a specific conversation by session ID or name.
-v, --version
Display version information.
--model _MODEL_
Specify the Claude model to use (aliases: sonnet, opus, haiku, or full model name).
--max-turns _N_
Maximum number of agentic turns in non-interactive mode.
--max-budget-usd _AMOUNT_
Maximum dollar amount for API calls (print mode only).
--output-format _FORMAT_
Output format: text, json, or stream-json.
--input-format _FORMAT_
Input format for print mode.
--tools _TOOLS_
Restrict available built-in tools (e.g., "Bash,Edit,Read").
--allowedTools _TOOLS_
Tools that execute without prompting.
--disallowedTools _TOOLS_
Tools removed from context.
--permission-mode _MODE_
Begin in specified permission mode (e.g., plan).
--system-prompt _PROMPT_
Replace the entire default system prompt.
--append-system-prompt _PROMPT_
Append additional instructions to the default prompt.
--add-dir _PATHS_
Add additional working directories.
--mcp-config _FILE_
Load MCP servers from JSON configuration file.
--chrome
Enable Chrome browser integration.
--ide
Auto-connect to IDE on startup.
--remote
Create web session on claude.ai.
--teleport
Resume web session in local terminal.
--verbose
Enable verbose logging output.
--debug _FILTER_
Enable debug mode with optional filtering.
--dangerously-skip-permissions
Skip permission prompts (use with caution in trusted environments).

FAQ

What is the claude command used for?

Claude Code is Anthropic's official command-line interface for interacting with Claude AI. It provides an agentic coding assistant that can read, write, and execute code directly in your development environment. The tool operates in two main modes: interactive mode for extended sessions with ongoing dialogue, and print mode (-p) for scripted automation and one-off commands. Claude can navigate codebases, edit files, run shell commands, and perform complex multi-step development tasks. Claude Code integrates with your terminal environment, understanding project context through file access and command execution. It maintains conversation history across sessions, allowing you to resume previous discussions. The tool supports piping input from other commands and outputting structured JSON for integration with other tools. Special syntax includes @ for file references and ! for direct shell command execution. The CLI supports MCP (Model Context Protocol) servers for extending capabilities and can connect to IDEs and Chrome for enhanced integration.

How do I run a basic claude example?

Run `claude` in a terminal, then adjust file names, paths, flags, or remote targets for your system.

What does -p, --print do in claude?

Print mode: process prompt and exit without interactive session.