# Migrating from the OpenAI API

> Keep the OpenAI SDK and reach Claude models by changing the base URL.

_Source: https://aiprimetech.io/docs/guides/migration-from-openai/ · Home > Docs > Working with the API_

The gateway exposes an OpenAI-compatible endpoint, so an existing OpenAI codebase can reach Claude models without rewriting request handling.


## Python

```python
from openai import OpenAI

client = OpenAI(
    api_key="sk-your-gateway-key",
    base_url="https://aiprimetech.io/v1",          # the /v1 is required here
)

resp = client.chat.completions.create(
    model="claude-sonnet-4-6",  # a Claude model through the OpenAI SDK
    messages=[{"role": "user", "content": "Hello"}],
)
```


## LangChain

```python
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="claude-sonnet-4-6",
    api_key="sk-your-gateway-key",
    base_url="https://aiprimetech.io/v1",
)
```


## Checklist

1. Set `base_url` to `https://aiprimetech.io/v1` — with the suffix.
2. Swap model ids to ones from [Models](/docs/getting-started/models/).
3. Response parsing is unchanged: `choices[0].message.content`.
4. Streaming is unchanged: `choices[0].delta.content` plus the `[DONE]` sentinel.
5. Function calling uses the OpenAI `tools` shape on this endpoint.

> Behaviour differs even when the interface does not. Claude and GPT respond differently to the same prompt, particularly on instruction-following and output formatting. Re-check prompts that depend on a specific style, and re-run evaluations rather than assuming parity.

- [Chat Completions](https://aiprimetech.io/docs/api-reference/chat-completions/) — Endpoint reference
- [Models](https://aiprimetech.io/docs/getting-started/models/) — Available ids
- [Base URLs](https://aiprimetech.io/docs/getting-started/base-urls/) — Why /v1 here


---
_ClaudeAPIKey.dev is an independently operated, Anthropic-compatible API gateway. Not affiliated with Anthropic._