AgentResult

Defined in: src/types/agent.ts:423

Result returned by the agent loop.

Constructors

Constructor

new AgentResult(data): AgentResult;

Defined in: src/types/agent.ts:477

Parameters

Parameter	Type
data	{ stopReason: StopReason; lastMessage: Message; invocationState: InvocationState; traces?: AgentTrace[]; metrics?: AgentMetrics; structuredOutput?: output<ZodType>; interrupts?: Interrupt[]; checkpoint?: Checkpoint; }
data.stopReason	StopReason
data.lastMessage	Message
data.invocationState	InvocationState
data.traces?	AgentTrace[]
data.metrics?	AgentMetrics
data.structuredOutput?	output<ZodType>
data.interrupts?	Interrupt[]
data.checkpoint?	Checkpoint

Returns

AgentResult

Properties

type

readonly type: "agentResult";

Defined in: src/types/agent.ts:424

---

stopReason

readonly stopReason: StopReason;

Defined in: src/types/agent.ts:429

The stop reason from the final model response.

---

lastMessage

readonly lastMessage: Message;

Defined in: src/types/agent.ts:434

The last message added to the messages array.

---

traces?

readonly optional traces?: AgentTrace[];

Defined in: src/types/agent.ts:440

Local execution traces collected during the agent invocation. Contains timing and hierarchy of operations within the agent loop.

---

structuredOutput?

readonly optional structuredOutput?: output<ZodType>;

Defined in: src/types/agent.ts:446

The validated structured output from the LLM, if a schema was provided. Type represents any validated Zod schema output.

---

metrics?

readonly optional metrics?: AgentMetrics;

Defined in: src/types/agent.ts:452

Aggregated metrics for the agent’s loop execution. Tracks cycle counts, token usage, tool execution stats, and model latency.

---

invocationState

readonly invocationState: InvocationState;

Defined in: src/types/agent.ts:460

Per-invocation state passed into the agent, threaded through hooks and tools, and surfaced here at the end of the invocation. See InvocationState for details. Always defined — defaults to {} when no invocationState was provided in InvokeOptions.

---

interrupts?

readonly optional interrupts?: Interrupt[];

Defined in: src/types/agent.ts:466

Interrupts that caused the agent to stop, when stopReason is 'interrupt'. Contains the unanswered interrupts that require human input to resume.

---

checkpoint?

readonly optional checkpoint?: Checkpoint;

Defined in: src/types/agent.ts:475

Experimental

Checkpoint captured when the agent paused for durable execution. Populated only when stopReason is 'checkpoint'. See the experimental checkpoint module for usage.

Accessors

contextSize

Get Signature

get contextSize(): number;

Defined in: src/types/agent.ts:512

The total prompt the model processed on the last invocation, including cached tokens. Convenience accessor that delegates to metrics.latestContextSize. Returns undefined when no metrics or invocations are available.

Returns

number

---

projectedContextSize

Get Signature

get projectedContextSize(): number;

Defined in: src/types/agent.ts:522

Projected context size for the next model call (total prompt including cached tokens plus the generated output from the last call). Convenience accessor that delegates to metrics.projectedContextSize. Returns undefined when no metrics or invocations are available.

Returns

number

Methods

toJSON()

toJSON(): object;

Defined in: src/types/agent.ts:536

Custom JSON serialization that excludes traces, metrics, and invocationState. Traces and metrics are excluded to avoid sending large payloads over the wire in API responses; invocationState is excluded because its values are caller-owned and may not be serializable (see InvocationState).

All three remain accessible via their properties for debugging.

Returns

object

Object representation for safe serialization

---

toString()

toString(): string;

Defined in: src/types/agent.ts:556

Extracts a string representation of the result.

Priority order:

1. interrupts serialized as JSON, if any are present
2. structuredOutput serialized as JSON
3. Text from textBlock, reasoningBlock, and citationsBlock content blocks

Returns

string

String representation of the result: JSON for interrupts/structuredOutput, or text content joined by newlines.