Implementation:CrewAIInc CrewAI CrewOutput Model

Knowledge Sources	CrewAI CrewAI Docs
Domains	Data_Processing, Multi_Agent_Systems
Last Updated	2026-02-11 00:00 GMT

Overview

Concrete data model for accessing crew execution results in multiple formats provided by the CrewAI framework.

Description

The CrewOutput class is a Pydantic model containing the final results of a crew execution. It provides access to the raw text output, optional Pydantic model, optional JSON dict, per-task outputs as a list of TaskOutput, and token usage metrics. It supports dictionary-style access via __getitem__ and string conversion.

Usage

Access CrewOutput fields after calling crew.kickoff(). Use .raw for text, .pydantic for typed model access, .json_dict for JSON, .tasks_output for per-task results, and .token_usage for cost tracking.

Code Reference

Source Location

• Repository: crewAI
• File: lib/crewai/src/crewai/crews/crew_output.py
• Lines: L13-61

Signature

class CrewOutput(BaseModel):
    """Class that represents the result of a crew."""

    raw: str = Field(description="Raw output of crew", default="")
    pydantic: BaseModel | None = Field(default=None)
    json_dict: dict[str, Any] | None = Field(default=None)
    tasks_output: list[TaskOutput] = Field(default=[])
    token_usage: UsageMetrics = Field(default_factory=UsageMetrics)

    @property
    def json(self) -> str | None: ...

    def to_dict(self) -> dict[str, Any]: ...

    def __getitem__(self, key): ...

    def __str__(self): ...

Import

from crewai.crews.crew_output import CrewOutput

I/O Contract

Inputs

Name	Type	Required	Description
raw	str	Yes	Raw text output from the crew (default: "")
pydantic	None	No	Pydantic model output if final task has output_pydantic
json_dict	None	No	JSON dict output if final task has output_json
tasks_output	list[TaskOutput]	No	Per-task output list
token_usage	UsageMetrics	No	Token consumption metrics

Outputs

Name	Type	Description
.raw	str	Raw text output
.pydantic	None	Typed Pydantic model
.json_dict	None	JSON dictionary
.json	None	Serialized JSON string
.tasks_output	list[TaskOutput]	Per-task outputs
.token_usage	UsageMetrics	Token usage (total_tokens, prompt_tokens, etc.)

Usage Examples

Accessing Results

result = crew.kickoff(inputs={"topic": "AI"})

# Raw text
print(result.raw)

# Structured output (if output_pydantic was set on final task)
if result.pydantic:
    report = result.pydantic
    print(report.title)

# JSON output (if output_json was set on final task)
if result.json_dict:
    print(result.json_dict["findings"])

# Per-task outputs
for task_output in result.tasks_output:
    print(f"Task: {task_output.description}")
    print(f"Result: {task_output.raw}")

# Token usage
print(f"Total tokens: {result.token_usage.total_tokens}")

Related Pages

Implements Principle

• Principle:CrewAIInc_CrewAI_Output_Processing