Anthropic Messages

POST /v1/messages speaks the Anthropic Messages format, so the Anthropic SDKs and tools built on them — Claude Code among them — work with only the base URL and key changed. Every text model answers on it, not only Claude.

A request

The Anthropic SDK appends /v1/messages itself, so its base URL has no /v1:

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

const client = new Anthropic({
  baseURL: "https://api.cloakray.com",
  apiKey: process.env.CLOAKRAY_API_KEY,
});

const message = await client.messages.create({
  model: "claude-opus-4-8",
  max_tokens: 1024,
  system: "You are a concise assistant.",
  messages: [{ role: "user", content: "Explain idempotency in one paragraph." }],
});

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

import os
import anthropic

client = anthropic.Anthropic(base_url="https://api.cloakray.com", api_key=os.environ["CLOAKRAY_API_KEY"])

message = client.messages.create(
    model="claude-opus-4-8",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain idempotency in one paragraph."}],
)
print(message.content[0].text)

curl https://api.cloakray.com/v1/messages \
  -H "x-api-key: $CLOAKRAY_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-4-8",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Explain idempotency in one paragraph."}]
  }'

The key goes in x-api-key or Authorization: Bearer; both work.

What carries over

The request and response follow the Anthropic format: max_tokens is required, system is a top-level field, and content is a list of typed blocks.

• Tools — tools with input_schema, tool_use blocks in the reply and tool_result blocks in your next turn. See tool calling.
• Images and PDFs — image and document blocks, as base64 or URL sources, on models that accept them. See vision and documents.
• Extended thinking — thinking with budget_tokens on models that reason. See reasoning.
• Prompt caching — cache_control on content blocks. See prompt caching.
• Streaming — stream: true returns the Anthropic event stream (message_start, content_block_delta, message_stop). See streaming.

Usage comes back in Anthropic's fields: input_tokens, output_tokens, cache_read_input_tokens and cache_creation_input_tokens.

Non-Claude models

Any text model can be called here — gpt-5.5, gemini-3.6-flash, kimi-k2.6 — which is what lets Claude Code and other Anthropic-native tools use them. Features a model doesn't have, such as extended thinking on a model that doesn't reason, are refused or ignored rather than imitated.

Errors

Errors come back in Anthropic's shape, {"type": "error", "error": {"type": "...", "message": "..."}}, so the SDK raises its usual exceptions. The statuses mean the same as everywhere else — see errors.