Interface: ResultMessage

Generated from the public TypeScript API.

Terminal event yielded at the end of every successful or failed run.

​

Extended by

• RunResult

​

Properties

​

durationMs

durationMs: number;

---

​

error?

optional error?: object;

​

cause?

optional cause?: unknown;

​

message

message: string;

---

​

messageId?

optional messageId?: string;

---

​

modelUsage?

optional modelUsage?: Record<string, ModelUsageSummary>;

---

​

parentMessageId?

optional parentMessageId?: string;

---

​

sessionId

sessionId: string;

---

​

stopReason?

optional stopReason?: string;

---

​

structuredOutput?

optional structuredOutput?: unknown;

---

​

subtype

subtype:
  | "success"
  | "error_aborted"
  | "error_during_execution"
  | "error_max_turns"
  | "error_max_budget"
  | "error_max_structured_output_retries";

---

​

text?

optional text?: string;

Assistant text from every text part of the turn’s final assistant message, identified by messageId.

Identical to the text on the assistant stream message with the same messageId, so this and the last streamed assistant text can never disagree.

undefined when the turn produced no assistant text (a tool-only final step, a result without a message id, or a hard failure before any matching assistant message was observed). Use text ?? "" for a plain string — which is exactly what RunResult.text does.

Unlike Claude Agent SDK’s result (success variant only), this is also populated on error subtypes: on error_aborted it is the partial answer produced before the interrupt.

---

​

totalCostUsd?

optional totalCostUsd?: number;

---

​

turns

turns: number;

---

​

usage?

optional usage?: UsageSummary;