Skip to main content
OpenAI provides GPT and o-series models. The AI Gateway supports OpenAI as a managed provider—no OpenAI account needed—and also works with your own OpenAI API key.

​
Managed keys setup

No OpenAI account needed. Follow the quickstart to create your gateway and API key, then point your SDK at the gateway: 
from openai import OpenAI

client = OpenAI(
    base_url="https://your-ai-gateway.ngrok.app/v1",
    api_key="ng-xxxxx-g1-xxxxx"  # Your AI Gateway API Key
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)

print(response.choices[0].message.content)

​
Attach your own key

Attach your OpenAI key to your AI Gateway API Key. ngrok encrypts and stores it server-side, so clients only ever send your AI Gateway API Key.
1

Get an OpenAI API key

Create an account at platform.openai.com and generate an API key. Keys start with sk-proj-.
2

Attach the key

ngrok api ai-gateway-provider-keys create \
  --ai-gateway-api-key-id aigk_xxxxx \
  --provider-id openai \
  --description "My OpenAI Key" \
  --value sk-proj-...
3

Make a request

Use your AI Gateway API Key—the gateway resolves the OpenAI key server-side.
from openai import OpenAI

client = OpenAI(
    base_url="https://your-ai-gateway.ngrok.app/v1",
    api_key="ng-xxxxx-g1-xxxxx"  # Your AI Gateway API Key
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)

print(response.choices[0].message.content)
See Attaching Provider Keys for key rotation, listing, and removing attached keys.

​
Explicit provider routing

Use the openai: prefix to force routing to OpenAI specifically: 
response = client.chat.completions.create(
    model="openai:gpt-4o",  # Always routes to OpenAI, not another provider
    messages=[{"role": "user", "content": "Hello!"}]
)
This is useful in multi-provider setups where the same model name might exist on multiple providers.

​
Available models

See the model catalog for the full list of supported models and their context windows.

​
Next steps