Get started

The Confidential Inference API is OpenAI-compatible: point an existing client at our base URL and keep the rest of your code. Versus non-confidential inference: about 1 to 3% token throughput overhead on B200s (benchmarks), negligible impact on Time to First Token.

1. Get an API key

API keys are provisioned by Confidential AI. Email hello@confidential.ai with the models you want, expected volume and use case, or use the request access form. DeepSeek V4.1 Flash is live; other models on request.

2. Make a request

The production base URL is https://api.confidential.ai. Call /v1/models to read the public model catalog, then call /v1/chat/completions with your API key:

curl https://api.confidential.ai/v1/chat/completions \
  -H "Authorization: Bearer $CONFIDENTIAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-V4.1-Flash",
    "messages": [{"role": "user", "content": "Explain remote attestation in one sentence."}]
  }'

Or with the OpenAI Python client:

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.confidential.ai/v1",
    api_key=os.environ["CONFIDENTIAL_API_KEY"],
)

resp = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V4.1-Flash",
    messages=[{"role": "user", "content": "Explain remote attestation in one sentence."}],
)
print(resp.choices[0].message.content)

Using a framework: OpenAI SDK, LangChain, Vercel AI SDK, LiteLLM.

Every endpoint, field, and error code is in the API reference.

3. Verify before you trust

The examples above use ordinary HTTPS. Before sending sensitive input, configure TEErminator with the reviewed release and deployment pins. Send requests through its local endpoint so it can verify the upstream connection before forwarding your data. Attestation links to the current setup guide. Overview explains how the pieces fit.