Skip to Content
Getting StartedPython Quickstart

Python Quickstart

The SDK configures the OpenAI client to route through the gateway automatically:

pip install curate-me openai
from curate_me.gateway import CurateGateway
 
gw = CurateGateway(api_key="cm_sk_xxx")
 
# Get a pre-configured OpenAI client pointed at the gateway
client = gw.openai()
 
response = client.chat.completions.create(
    model="anthropic/claude-haiku-4-5",
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

You can also pass a provider key explicitly if you are not using stored secrets:

client = gw.openai(provider_key="sk-your-openai-key")

Using the OpenAI SDK Directly

If you prefer not to use the Curate-Me SDK, just change the base URL and point the OpenAI SDK at the gateway. Set api_key to your Curate-Me gateway key (cm_sk_...) — the SDK forwards it as Authorization: Bearer cm_sk_..., which the gateway accepts, and your provider key is resolved from your stored Provider Secret. Do not put a placeholder string here: any non-cm_sk_ value is treated as an inline provider key and forwarded upstream, which fails with a 401.

from openai import OpenAI
 
client = OpenAI(
    base_url="https://api.curate-me.ai/v1/openrouter",
    api_key="cm_sk_xxx",  # your Curate-Me gateway key
    default_headers={"X-CM-API-Key": "cm_sk_xxx"},
)
 
response = client.chat.completions.create(
    model="anthropic/claude-haiku-4-5",
    messages=[{"role": "user", "content": "Explain AI governance in one sentence."}],
)
print(response.choices[0].message.content)

To send a provider key inline instead of storing it, add an X-Provider-Key header rather than overloading api_key:

client = OpenAI(
    base_url="https://api.curate-me.ai/v1/openrouter",
    api_key="cm_sk_xxx",
    default_headers={
        "X-CM-API-Key": "cm_sk_xxx",
        "X-Provider-Key": "sk-your-openrouter-key",
    },
)

Using Anthropic Models (Native SDK)

from curate_me.gateway import CurateGateway
 
gw = CurateGateway(api_key="cm_sk_xxx")
client = gw.anthropic()
 
message = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
)
print(message.content[0].text)

Checking Usage

# Response headers include governance metadata
print(response.headers.get("X-CM-Cost"))
print(response.headers.get("X-CM-Request-Id"))
print(response.headers.get("RateLimit-Remaining"))
print(response.headers.get("X-CM-Governance-Time-Ms"))
InstallationTypeScript Quickstart