# SDKs and clients

> Use the OpenAI and Anthropic SDKs, the Vercel AI SDK, LiteLLM and LangChain by changing the base URL.

Deference has no SDK of its own. The SDKs you already use work as they are: set the base URL and the key.

| Client                             | Base URL                  |
| ---------------------------------- | ------------------------- |
| OpenAI SDK (Python, TypeScript)    | `https://deference.si/v1` |
| Anthropic SDK (Python, TypeScript) | `https://deference.si`    |
| Vercel AI SDK, LiteLLM, LangChain  | `https://deference.si/v1` |

The examples read the key from `DEFERENCE_API_KEY`.

## OpenAI SDK

**Python**

```python
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://deference.si/v1",
    api_key=os.environ["DEFERENCE_API_KEY"],
)

response = client.chat.completions.create(
    model="anthropic/claude-sonnet-5.5",
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)
print(response.choices[0].message.content)
```

**TypeScript**

```typescript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://deference.si/v1",
  apiKey: process.env.DEFERENCE_API_KEY,
});

const response = await client.chat.completions.create({
  model: "anthropic/claude-sonnet-5.5",
  messages: [{ role: "user", content: "Say hello in one sentence." }],
});
console.log(response.choices[0].message.content);
```

`client.responses` and `client.embeddings` work the same way. The SDK's image methods call OpenAI's own image paths, which Deference does not serve. Call [`POST /v1/images`](https://deference.si/docs/api-reference/images) directly.

## Anthropic SDK

The Anthropic SDK adds `/v1/messages` itself, so its base URL has no `/v1`. It sends the key as `x-api-key`.

**Python**

```python
import os
import anthropic

client = anthropic.Anthropic(
    base_url="https://deference.si",
    api_key=os.environ["DEFERENCE_API_KEY"],
)

message = client.messages.create(
    model="anthropic/claude-sonnet-5.5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)
print(message.content[0].text)
```

**TypeScript**

```typescript
import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "https://deference.si",
  apiKey: process.env.DEFERENCE_API_KEY,
});

const message = await client.messages.create({
  model: "anthropic/claude-sonnet-5.5",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Say hello in one sentence." }],
});
console.log(message.content);
```

`messages.count_tokens` is not available. The Claude Agent SDK reads `ANTHROPIC_BASE_URL` and `ANTHROPIC_AUTH_TOKEN`, as [Claude Code](https://deference.si/docs/coding-tools/claude-code) does.

## Vercel AI SDK

Use the OpenAI-compatible provider.

```typescript
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { generateText } from "ai";

const deference = createOpenAICompatible({
  name: "deference",
  baseURL: "https://deference.si/v1",
  apiKey: process.env.DEFERENCE_API_KEY,
});

const { text } = await generateText({
  model: deference("anthropic/claude-sonnet-5.5"),
  prompt: "Say hello in one sentence.",
});
console.log(text);
```

## LiteLLM

Prefix the model with `openai/` so LiteLLM uses the OpenAI-compatible path, then add Deference's id.

```python
import os
import litellm

response = litellm.completion(
    model="openai/anthropic/claude-sonnet-5.5",
    api_base="https://deference.si/v1",
    api_key=os.environ["DEFERENCE_API_KEY"],
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)
print(response.choices[0].message.content)
```

## LangChain

```python
import os
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="anthropic/claude-sonnet-5.5",
    base_url="https://deference.si/v1",
    api_key=os.environ["DEFERENCE_API_KEY"],
)
print(llm.invoke("Say hello in one sentence.").content)
```

## Plain HTTP

No SDK is needed. Every endpoint takes JSON and a bearer key. See [Chat completions](https://deference.si/docs/api-reference/chat-completions) for a curl example.

Setting up a coding tool instead? See [Coding tools](https://deference.si/docs/coding-tools).