> ## Documentation Index
> Fetch the complete documentation index at: https://agno-v2-codex-docs-audit-20260719-0149.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent-as-Judge Eval Metrics

> Attach AgentAsJudgeEval as an agent post_hook and read the evaluator's token usage from run_output.metrics.details['eval_model'].

Demonstrates that eval model metrics are accumulated back into the original agent's run\_output when AgentAsJudgeEval is used as a post\_hook.

```python agent_as_judge_eval_metrics.py theme={null}
"""
Agent-as-Judge Eval Metrics
============================

Demonstrates that eval model metrics are accumulated back into the
original agent's run_output when AgentAsJudgeEval is used as a post_hook.

After the agent runs, the evaluator agent makes its own model call.
Those eval tokens show up under "eval_model" in run_output.metrics.details.
"""

from agno.agent import Agent
from agno.eval.agent_as_judge import AgentAsJudgeEval
from agno.models.openai import OpenAIChat
from rich.pretty import pprint

# ---------------------------------------------------------------------------
# Create eval as a post-hook
# ---------------------------------------------------------------------------
eval_hook = AgentAsJudgeEval(
    name="Quality Check",
    model=OpenAIChat(id="gpt-4o-mini"),
    criteria="Response should be accurate, clear, and concise",
    scoring_strategy="binary",
)

agent = Agent(
    model=OpenAIChat(id="gpt-4o-mini"),
    instructions="Answer questions concisely.",
    post_hooks=[eval_hook],
)

# ---------------------------------------------------------------------------
# Run
# ---------------------------------------------------------------------------
if __name__ == "__main__":
    result = agent.run("What is the capital of France?")

    # The run metrics now include both agent model + eval model tokens
    if result.metrics:
        print("Total tokens (agent + eval):", result.metrics.total_tokens)

        if result.metrics.details:
            # Agent's own model call
            if "model" in result.metrics.details:
                agent_tokens = sum(
                    metric.total_tokens for metric in result.metrics.details["model"]
                )
                print("Agent model tokens:", agent_tokens)

            # Eval model call (accumulated from evaluator agent)
            if "eval_model" in result.metrics.details:
                eval_tokens = sum(
                    metric.total_tokens
                    for metric in result.metrics.details["eval_model"]
                )
                print("Eval model tokens:", eval_tokens)
                for metric in result.metrics.details["eval_model"]:
                    print(f"  Evaluator: {metric.id} ({metric.provider})")

            print("\nFull metrics details:")
            pprint(result.metrics.to_dict())
```

## Run the Example

<Steps>
  <Snippet file="create-venv-step.mdx" />

  <Step title="Install dependencies">
    ```bash theme={null}
    uv pip install -U agno openai
    ```
  </Step>

  <Step title="Export your OpenAI API key">
    <CodeGroup>
      ```bash Mac/Linux theme={null}
      export OPENAI_API_KEY="your_openai_api_key_here"
      ```

      ```bash Windows theme={null}
      $Env:OPENAI_API_KEY="your_openai_api_key_here"
      ```
    </CodeGroup>
  </Step>

  <Step title="Run the example">
    Save the code above as `agent_as_judge_eval_metrics.py`, then run:

    ```bash theme={null}
    python agent_as_judge_eval_metrics.py
    ```
  </Step>
</Steps>

Full source: [cookbook/09\_evals/agent\_as\_judge/agent\_as\_judge\_eval\_metrics.py](https://github.com/agno-agi/agno/blob/v2.7.4/cookbook/09_evals/agent_as_judge/agent_as_judge_eval_metrics.py)