Agent SDKInstall

Summary

Continue a previous conversation with the Claude Agent SDK using specific prompts and options.

Inputs

• prompt (required) — The prompt or instruction to send to Claude Agent SDK.
• model — Claude model to use, either 'Sonnet' or 'Opus'.
• maxTurns — Maximum number of conversation turns allowed.
• timeout — Maximum time to wait for completion before aborting.
• projectPath — Directory path where Claude Agent SDK should run.
• outputFormat — Format of the output data: 'Structured', 'Messages', or 'Text'.
• allowedTools — Select which built-in tools are allowed during execution.
• disallowedTools — Select which built-in tools are explicitly blocked.
• additionalOptions — Additional options like debug mode, fallback model, and permission mode.

Output shape

a structured object containing conversation details, results, and metrics

The output includes the conversation messages, a summary, the final result, and additional metrics. Specific formatting based on the selected outputFormat option.