CLI Flags
Session management
| Flag | Purpose |
|---|---|
--continue | Resume the most recent session |
--resume [id] | Resume a specific session by ID, or open a picker |
--fork-session | Branch off from a resumed session without modifying the original (use with --resume or --continue) |
--from-pr [number] | Start a session from a GitHub pull request |
--session-id <uuid> | Use a specific session ID (must be a valid UUID) |
Model and effort
| Flag | Purpose |
|---|---|
--model sonnet | Use a specific model (aliases like sonnet, opus, or full model IDs) |
--fallback-model <model> | Automatic fallback when the primary model is overloaded (headless mode only) |
--effort <level> | Set reasoning depth: low, medium, or high |
--max-budget-usd 5 | Set a spending cap for the session (headless mode only) |
--max-turns 20 | Limit the number of agentic turns |
Permissions and tool control
| Flag | Purpose |
|---|---|
--permission-mode <mode> | Permission mode: default, acceptEdits, plan, bypassPermissions, dontAsk |
--allowedTools <tools> | Allow specific tools (e.g. "Bash(git:*) Edit Read") |
--disallowedTools <tools> | Deny specific tools (e.g. "Bash(rm:*)") |
--tools <tools> | Restrict the available tool set entirely ("" to disable all, "default" for all) |
--dangerously-skip-permissions | Bypass all permission checks (sandboxed environments only) |
warning
--dangerously-skip-permissions disables all safety prompts. Claude Code will read, write, and execute anything without asking. Only use this inside a fully sandboxed environment with no internet access and no access to sensitive data (e.g. a disposable Docker container or CI job). Never use it on your main development machine. Prefer scoped --allowedTools rules to grant exactly the access you need instead.
Headless and scripting
| Flag | Purpose |
|---|---|
--print "prompt" | Run a single prompt in headless mode (no interactive session) |
--output-format <format> | Output format with --print: text, json, or stream-json |
--input-format <format> | Input format with --print: text or stream-json |
--json-schema <schema> | Enforce structured output with a JSON Schema |
--no-session-persistence | Do not save the session to disk (headless mode only) |
System prompts
| Flag | Purpose |
|---|---|
--system-prompt <prompt> | Replace the default system prompt entirely |
--append-system-prompt <prompt> | Append instructions to the default system prompt |
Agents and plugins
| Flag | Purpose |
|---|---|
--agent <agent> | Use a named agent for the session |
--agents <json> | Define custom agents inline as JSON |
--plugin-dir <paths> | Load plugins from directories for this session |
MCP servers
| Flag | Purpose |
|---|---|
--mcp-config <config> | Load MCP servers from a JSON file or string |
--strict-mcp-config | Only use MCP servers from --mcp-config, ignore all other config |
Files and directories
| Flag | Purpose |
|---|---|
--add-dir <directories> | Grant tool access to directories outside the working directory |
--worktree [name] | Create a new git worktree for isolated work |
--tmux | Create a tmux session for the worktree (requires --worktree) |
Debug
| Flag | Purpose |
|---|---|
--debug [filter] | Enable debug mode with optional category filtering (e.g. "api,hooks") |
--verbose | Enable verbose output |
Environment variables
| Variable | Purpose |
|---|---|
CLAUDE_CODE_EFFORT_LEVEL | Set effort level: high, medium, or low |
MAX_THINKING_TOKENS | Control how many tokens the model uses for reasoning |