Anthropic Format

LLMoxy provides a fully compatible Anthropic Messages API interface, optimized for the Claude series models. Users of the official Anthropic SDK only need to change base_url to switch.

Base URL

https://llmoxy.com/v1

Authentication

Two authentication methods are supported:

Method 1: x-api-key (recommended)

x-api-key: YOUR_API_KEY

Method 2: Authorization

Authorization: Bearer YOUR_API_KEY

Messages API

Claude's core conversation interface.

Request URL

POST /v1/messages

Request Example

curl https://llmoxy.com/v1/messages \
-H "content-type: application/json" \
-H "x-api-key: YOUR_API_KEY" \
-d '{
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"messages": [
{
"role": "user",
"content": "Write a binary search algorithm in Python"
}
]
}'

Request Parameters

Message Format

Unlike OpenAI, the Anthropic format uses a separate parameter for the system prompt:

{
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"system": "You are a Python programming expert",
"messages": [
{
"role": "user",
"content": "Write a quicksort"
},
{
"role": "assistant",
"content": "Sure, I'll write a Python quicksort..."
},
{
"role": "user",
"content": "Add detailed comments"
}
]
}

Note

• messages can only contain user and assistant roles
• Must start with a user message
• user and assistant messages must alternate

Response Format

{
"id": "msg_123",
"type": "message",
"role": "assistant",
"content": [
{
"type": "text",
"text": "This is Claude's reply content"
}
],
"model": "claude-sonnet-4-20250514",
"stop_reason": "end_turn",
"usage": {
"input_tokens": 25,
"output_tokens": 150
}
}

Streaming Output

Set stream: true to enable streaming:

curl https://llmoxy.com/v1/messages \
-H "content-type: application/json" \
-H "x-api-key: YOUR_API_KEY" \
-d '{
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'

Streaming Response Format

event: message_start
data: {"type":"message_start","message":{"id":"msg_123",...}}

event: content_block_delta
data: {"type":"content_block_delta","delta":{"type":"text_delta","text":"H"}}

event: content_block_delta
data: {"type":"content_block_delta","delta":{"type":"text_delta","text":"i"}}

event: message_stop
data: {"type":"message_stop"}

Supported Models

TIP See the official model hub for currently available models.

Multimodal Support

Claude supports image input:

{
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"messages": [
{
"role": "user",
"content": [
{
"type": "image",
"source": {
"type": "base64",
"media_type": "image/jpeg",
"data": "/9j/4AAQSkZJRg..."
}
},
{
"type": "text",
"text": "What is in this image?"
}
]
}
]
}

Supported image formats: JPEG, PNG, GIF, WebP. Max 5MB per image.

Code Examples

Python (Official SDK)

from anthropic import Anthropic

client = Anthropic(
api_key="YOUR_API_KEY",
base_url="https://llmoxy.com/v1"
)

message = client.messages.create(
model="claude-sonnet-4-20250514",
max_tokens=1024,
messages=[
{"role": "user", "content": "Write a Python decorator"}
]
)

print(message.content[0].text)

Python (Streaming)

with client.messages.stream(
model="claude-sonnet-4-20250514",
max_tokens=1024,
messages=[{"role": "user", "content": "Write an article"}]
) as stream:
for text in stream.text_stream:
print(text, end="", flush=True)

Node.js (Official SDK)

import Anthropic from '@anthropic-ai/sdk';

const client = new Anthropic({
apiKey: 'YOUR_API_KEY',
baseURL: 'https://llmoxy.com/v1'
});

const message = await client.messages.create({
model: 'claude-sonnet-4-20250514',
max_tokens: 1024,
messages: [
{ role: 'user', content: 'Write a quicksort' }
]
});

console.log(message.content[0].text);

cURL

curl https://llmoxy.com/v1/messages \
-H "content-type: application/json" \
-H "x-api-key: YOUR_API_KEY" \
-d '{
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"messages": [
{"role": "user", "content": "Hello, Claude!"}
]
}'

Error Handling

Error Response Format

{
"type": "error",
"error": {
"type": "invalid_request_error",
"message": "Error description"
}
}

Common Error Types

Error Handling Example

from anthropic import Anthropic, APIError

client = Anthropic(
api_key="YOUR_API_KEY",
base_url="https://llmoxy.com/v1"
)

try:
message = client.messages.create(
model="claude-sonnet-4-20250514",
max_tokens=1024,
messages=[{"role": "user", "content": "Hello"}]
)
except APIError as e:
print(f"Error type: {e.type}")
print(f"Error message: {e.message}")

Advanced Usage

1. System Prompt

message = client.messages.create(
model="claude-sonnet-4-20250514",
max_tokens=1024,
system="You are a Python expert, good at data analysis and machine learning.",
messages=[
{"role": "user", "content": "How to handle missing values with pandas?"}
]
)

2. Multi-turn Conversation

conversation = [
{"role": "user", "content": "Write a user class"},
{"role": "assistant", "content": "Sure, I'll write a User class..."},
{"role": "user", "content": "Add password encryption"}
]

message = client.messages.create(
model="claude-sonnet-4-20250514",
max_tokens=1024,
messages=conversation
)

3. Image Analysis

import base64

with open("image.jpg", "rb") as f:
image_data = base64.b64encode(f.read()).decode()

message = client.messages.create(
model="claude-sonnet-4-20250514",
max_tokens=1024,
messages=[
{
"role": "user",
"content": [
{
"type": "image",
"source": {
"type": "base64",
"media_type": "image/jpeg",
"data": image_data
}
},
{
"type": "text",
"text": "Analyze this architecture diagram"
}
]
}
]
)

Best Practices

1. Set max_tokens Reasonably

Different models have different maximum outputs: Opus 32K, Sonnet 64K, Haiku 8K. Set according to actual needs to avoid waste.

2. Use the system Parameter for Prompts

Put role definitions in the system parameter, not in messages:

# ✅ Recommended
message = client.messages.create(
system="You are a Python expert",
messages=[{"role": "user", "content": "Write code"}],
...
)

# ❌ Not recommended
message = client.messages.create(
messages=[{"role": "user", "content": "You are a Python expert. Write code"}],
...
)

3. Handle Long Conversations

Claude has a 200K context window, but it is recommended to periodically summarize conversations and remove unimportant historical messages to save token consumption.

4. Implement Retry Logic

from tenacity import retry, stop_after_attempt, wait_exponential

@retry(
stop=stop_after_attempt(3),
wait=wait_exponential(multiplier=1, min=4, max=10)
)
def call_claude():
return client.messages.create(...)

Next Steps

• Groups and Pricing — learn about pricing details
• Get Support — contact us if you encounter issues