Use Anthropic models through the OpenAI SDK

Anthropic's API has a slightly different shape than OpenAI's — different system prompt position, different tool schema, different streaming events. Most teams that started on OpenAI either install both SDKs and write an adapter, or block on a rewrite.

You don't have to. Point the OpenAI SDK at AIgateway and Claude is a model slug change.

Switch to Claude Opus 4.7

from openai import OpenAI

client = OpenAI(
    base_url="https://api.aigateway.sh/v1",
    api_key="sk-aig-...",
)

resp = client.chat.completions.create(
    model="anthropic/claude-opus-4.7",
    messages=[
        {"role": "system", "content": "You are a careful editor."},
        {"role": "user",   "content": "Tighten this paragraph: ..."},
    ],
)

Tool calling, vision, JSON mode

All three pass through. The OpenAI tool-call shape (name + arguments JSON) is translated to Anthropic's tool_use blocks server-side; the response comes back in OpenAI shape so your existing parser keeps working.

Vision: send image_url parts in the messages array exactly like you would for GPT-5.4. We translate to Anthropic's image content blocks.

JSON mode: response_format={"type": "json_object"} is honored.

Prompt caching

Anthropic supports automatic prompt caching for messages above a length threshold. AIgateway proxies the cache_control hints through, and bills cache reads at 10% of the uncached cost. If your prompts repeat the same long context, cache hit rates over 30% are realistic.

Want to keep using the Anthropic SDK?

If you've already written code against the Anthropic SDK, point it at our Anthropic-shape endpoint instead. Same models, same shape, no other changes.

import anthropic

client = anthropic.Anthropic(
    base_url="https://api.aigateway.sh/anthropic",
    api_key="sk-aig-...",
)
client.messages.create(
    model="claude-opus-4.7",
    max_tokens=1024,
    messages=[{"role": "user", "content": "hi"}],
)

Related guides