Portkey

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

Get a Portkey API key from your workspace.

Create a provider integration in your Portkey Model Catalog, then copy an allowed model string in the form @provider_slug/model_name. Replace @your-provider-slug/gpt-5-nano in the example with that value. The slug is specific to your workspace. These catalog examples need PORTKEY_API_KEY; they do not require a separate virtual key. Pass the key explicitly as portkey_api_key, because the Agno adapter does not read that environment variable automatically.

Legacy virtual-key routing is also supported: configure virtual_key (or PORTKEY_VIRTUAL_KEY) when using that route. It is an alternative to the catalog model slug.

export PORTKEY_API_KEY=***

Example

Install the portkey-ai and openai packages:

uv pip install -U agno portkey-ai openai

Save the following as agent.py and run python agent.py:

import os

from agno.agent import Agent
from agno.models.portkey import Portkey

agent = Agent(
    model=Portkey(
        id="@your-provider-slug/gpt-5-nano",
        portkey_api_key=os.getenv("PORTKEY_API_KEY"),
    ),
    markdown=True
)

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

Advanced Configuration

For cross-provider fallback, configure each target with its own catalog model string. Replace both provider slugs below with providers configured in your workspace, and use models allowed by those integrations:

import os

from agno.agent import Agent
from agno.models.portkey import Portkey

config = {
    "strategy": {
        "mode": "fallback"
    },
    "targets": [
        {"override_params": {"model": "@your-openai-provider/gpt-5-mini"}},
        {"override_params": {"model": "@your-anthropic-provider/claude-sonnet-4-6"}}
    ]
}

agent = Agent(
    model=Portkey(
        id="@your-openai-provider/gpt-5-mini",
        config=config,
        portkey_api_key=os.getenv("PORTKEY_API_KEY"),
    ),
)

Params

Portkey also supports the params of OpenAI.