> ## Documentation Index
> Fetch the complete documentation index at: https://docs-v1.agno.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Portkey

> Learn how to use models through the Portkey AI Gateway in Agno.

Portkey is an AI Gateway that provides a unified interface to access multiple AI providers with advanced features like routing, load balancing, retries, and observability. Use Portkey to build production-ready AI applications with better reliability and cost optimization.

With Portkey, you can:

* Route requests across multiple AI providers
* Implement fallback mechanisms for better reliability
* Monitor and analyze your AI usage
* Cache responses for cost optimization
* Apply rate limiting and usage controls

## Authentication

You need both a Portkey API key and a virtual key for model routing. Get them [from Portkey here](https://app.portkey.ai/).

<CodeGroup>
  ```bash Mac theme={null}
  export PORTKEY_API_KEY=***
  export PORTKEY_VIRTUAL_KEY=***
  ```

  ```bash Windows theme={null}
  setx PORTKEY_API_KEY ***
  setx PORTKEY_VIRTUAL_KEY ***
  ```
</CodeGroup>

## Example

Use `Portkey` with your `Agent`:

<CodeGroup>
  ```python agent.py theme={null}
  from agno.agent import Agent, RunResponse
  from agno.models.portkey import Portkey

  agent = Agent(
      model=Portkey(id="gpt-4o-mini"),
      markdown=True
  )

  # Print the response in the terminal
  agent.print_response("What is Portkey and why would I use it as an AI gateway?")
  ```
</CodeGroup>

## Advanced Configuration

You can configure Portkey with custom routing and retry policies:

```python theme={null}
from agno.agent import Agent
from agno.models.portkey import Portkey

config = {
    "strategy": {
        "mode": "fallback"
    },
    "targets": [
        {"virtual_key": "openai-key"},
        {"virtual_key": "anthropic-key"}
    ]
}

agent = Agent(
    model=Portkey(
        id="gpt-4o-mini",
        config=config,
    ),
)
```

<Note> View more examples [here](../examples/models/portkey). </Note>

## Params

<Snippet file="model-portkey-params.mdx" />

`Portkey` also supports the params of [OpenAI](/reference/models/openai).
