Claude Messages

The native endpoint uses POST /v1/messages, x-api-key, and anthropic-version. Its body and response differ from OpenAI Chat Completions.

curl

# Claude native authentication uses x-api-key, not Authorization: Bearer.
curl https://sprelaytoken.com/v1/messages \
  -H "x-api-key: $SPRELAY_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 512,
    "messages": [
      {"role": "user", "content": "Explain what an API relay does."}
    ]
  }'

anthropic-version selects the protocol version. max_tokens is required and caps output length. JSON cannot contain comments, so preserve all quotes, commas, and brackets.

Python SDK

# Import the Anthropic SDK and read the key from the environment.
import os
from anthropic import Anthropic

# The Claude SDK appends /v1/messages to the root URL.
client = Anthropic(
    api_key=os.environ["SPRELAY_API_KEY"],
    base_url="https://sprelaytoken.com",
)

# max_tokens is required by the Messages API.
message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=512,
    messages=[
        {"role": "user", "content": "Explain what an API relay does."}
    ],
)

# A plain-text response is stored in the first content block.
print(message.content[0].text)

JavaScript SDK

// Import the Anthropic SDK after installing @anthropic-ai/sdk.
import Anthropic from '@anthropic-ai/sdk'

const client = new Anthropic({
  apiKey: process.env.SPRELAY_API_KEY,
  baseURL: 'https://sprelaytoken.com',
})

// The SDK builds the native /v1/messages request.
const message = await client.messages.create({
  model: 'claude-sonnet-4-6',
  max_tokens: 512,
  messages: [{ role: 'user', content: 'Explain what an API relay does.' }],
})

// content is an array of typed blocks; inspect each type in production.
console.log(message.content)

Claude system instructions normally use the top-level system field, not a system role inside messages. The response uses content blocks instead of OpenAI choices.