Struct ClaudeAgentOptions 

pub struct ClaudeAgentOptions {

    pub prompt: Option<String>,
    pub model: Option<String>,
    pub system_prompt: Option<String>,
    pub append_system_prompt: Option<String>,
    pub max_turns: Option<u32>,
    pub max_tokens: Option<u32>,
    pub session_id: Option<String>,
    pub continue_session: bool,
    pub cwd: Option<PathBuf>,
    pub permission_mode: PermissionMode,
    pub allowed_tools: Vec<String>,
    pub can_use_tool: Option<CanUseToolCallback>,
    pub hooks: Vec<HookDefinition>,
    pub mcp_servers: HashMap<String, McpServerConfig>,
    pub agents: Vec<AgentDefinition>,
    pub sandbox: Option<SandboxSettings>,
    pub env: HashMap<String, String>,
    pub verbose: bool,
    pub cli_path: Option<PathBuf>,
    pub extra_cli_args: Vec<String>,
    pub connect_timeout: Option<Duration>,
    pub control_timeout: Option<Duration>,
    pub on_stderr: Option<StderrCallback>,
    pub no_cache: bool,
    pub temperature: Option<f64>,
    pub context_window: Option<f64>,
}

Configuration options for a Claude Agent SDK query or client.

All fields are public with sensible defaults. Use ..Default::default() for fields you don’t need to set.

Fields

prompt: Option<String>

The prompt/message to send. Can be set here or passed to query().

model: Option<String>

Model to use (e.g. “claude-sonnet-4-20250514”).

system_prompt: Option<String>

System prompt override.

append_system_prompt: Option<String>

Append system prompt (added after default).

max_turns: Option<u32>

Maximum turns (agentic loops) before stopping.

max_tokens: Option<u32>

Maximum tokens in the response.

session_id: Option<String>

Resume an existing session by ID.

continue_session: bool

Continue the most recent session.

cwd: Option<PathBuf>

Working directory for the CLI process.

permission_mode: PermissionMode

Permission mode for tool usage.

allowed_tools: Vec<String>

Specific tools to allow (when using AllowedTools mode).

can_use_tool: Option<CanUseToolCallback>

Custom permission callback.

hooks: Vec<HookDefinition>

Registered hook definitions.

mcp_servers: HashMap<String, McpServerConfig>

MCP servers to register with the CLI.

agents: Vec<AgentDefinition>

Sub-agent definitions.

sandbox: Option<SandboxSettings>

Sandbox configuration.

env: HashMap<String, String>

Additional environment variables for the CLI process.

verbose: bool

Verbose output from CLI.

cli_path: Option<PathBuf>

Path to the claude CLI binary (auto-detected if None).

extra_cli_args: Vec<String>

Custom CLI arguments (appended after built-in ones).

connect_timeout: Option<Duration>

Timeout for the initial connection handshake.

control_timeout: Option<Duration>

Timeout for control protocol requests.

on_stderr: Option<StderrCallback>

Stderr callback - receives stderr lines from CLI process.

no_cache: bool

Disallow use of the prompt cache

temperature: Option<f64>

Temperature setting

context_window: Option<f64>

Context window fraction (0.0-1.0) to use before summarizing.

Implementations

impl ClaudeAgentOptions

pub fn resolve_cli_path(&self) -> Result<PathBuf>

Resolve the CLI path: use the configured one or auto-detect.

Trait Implementations

impl Debug for ClaudeAgentOptions

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for ClaudeAgentOptions

fn default() -> ClaudeAgentOptions

Returns the “default value” for a type.