LLM — API reference

call_claude

call_claude(
    prompt: str,
    *,
    system_prompt: str | None = None,
    model: str = "haiku",
    timeout: int | None = None,
    retries: int = 0,
    backoff_base: float = 2.0,
) -> str | None

Invoke claude -p --model <model> with the prompt via stdin.

PARAMETER	DESCRIPTION
prompt	The user prompt to send. TYPE: str
system_prompt	Optional system prompt prepended with a marker. TYPE: str | None DEFAULT: None
model	Model name passed to --model (default: haiku). TYPE: str DEFAULT: 'haiku'
timeout	Subprocess timeout in seconds. None (the default) reads the llm-timeout setting. TYPE: int | None DEFAULT: None
retries	Number of retries on transient failure (default: 0). TYPE: int DEFAULT: 0
backoff_base	Base for exponential backoff in seconds (default: 2.0). TYPE: float DEFAULT: 2.0

RETURNS	DESCRIPTION
str | None	Stripped stdout on success, or None on failure.