Messages API

Create Claude conversations through the Anthropic native protocol. HaoAI is fully compatible with the Anthropic Messages API, so you can use the official SDK directly.

Endpoint


POST https://api.hao.ai/anthropic/v1/messages

Authentication

The Anthropic protocol uses the x-api-key header:


x-api-key: <your HAOAI_API_KEY>
anthropic-version: 2023-06-01

Request Parameters

Parameter	Type	Required	Description
`model`	string	✅	Model identifier, e.g. `anthropic/claude-sonnet-4.6`
`max_tokens`	number	✅	Maximum tokens to generate
`messages`	array	✅	Message array
`system`	string	—	System prompt
`temperature`	number	—	Sampling temperature 0-1
`top_p`	number	—	Nucleus sampling parameter
`top_k`	number	—	Top-K sampling
`stream`	boolean	—	Enable streaming response
`tools`	array	—	Tool definitions
`tool_choice`	object	—	Tool selection strategy
`thinking`	object	—	Claude thinking configuration; valid values depend on the model version
`output_config`	object	—	Output configuration for Claude 4.6+ adaptive thinking, such as `effort`

Message Format


interface Message {
  role: 'user' | 'assistant'
  content: string | ContentBlock[]
}

type ContentBlock =
  | { type: 'text'; text: string }
  | { type: 'image'; source: { type: 'base64'; media_type: string; data: string } }
  | { type: 'tool_use'; id: string; name: string; input: object }
  | { type: 'tool_result'; tool_use_id: string; content: string }

Claude Thinking Modes

Claude thinking parameters differ by model version. Configure thinking for the exact model you call, and do not send adaptive or budget_tokens to models that do not support them.

Model	Recommended configuration	Notes
`anthropic/claude-sonnet-4.5`, `anthropic/claude-haiku-4.5`	Manual thinking	Use `enabled + budget_tokens`; does not support `adaptive`
`anthropic/claude-sonnet-4.6`, `anthropic/claude-opus-4.6`	Adaptive thinking	Use `adaptive`; manual `budget_tokens` still works, but is no longer recommended
`anthropic/claude-opus-4.7+`	Adaptive thinking	Use `adaptive` only; does not support manual `budget_tokens`

budget_tokens must be less than max_tokens. Thinking consumes output tokens and is billed as output usage. If you only need the final answer, use display: "omitted" where supported to avoid returning thinking text.

Manual Thinking: Sonnet 4.5 / Haiku 4.5

claude-sonnet-4.5 and claude-haiku-4.5 use manual thinking and require an explicit budget_tokens value.


{
  "model": "anthropic/claude-sonnet-4.5",
  "max_tokens": 16000,
  "thinking": {
    "type": "enabled",
    "budget_tokens": 10000
  },
  "messages": [
    {
      "role": "user",
      "content": "Analyze this complex problem in depth..."
    }
  ]
}

Adaptive Thinking: Sonnet 4.6 / Opus 4.6

claude-sonnet-4.6 and claude-opus-4.6 should use adaptive thinking. Set output_config.effort to low, medium, or high based on task complexity.


{
  "model": "anthropic/claude-sonnet-4.6",
  "max_tokens": 16000,
  "thinking": {
    "type": "adaptive"
  },
  "output_config": {
    "effort": "medium"
  },
  "messages": [
    {
      "role": "user",
      "content": "Analyze this complex problem in depth..."
    }
  ]
}

Opus 4.7+: Do Not Use budget_tokens

claude-opus-4.7+ should continue to use thinking: {"type":"adaptive"}. Do not send manual budget_tokens to these models, or the upstream provider may return a parameter error.

Request Examples

cURL

Terminal


curl https://api.hao.ai/anthropic/v1/messages \
  -H "x-api-key: $HAOAI_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-4.6",
    "max_tokens": 1024,
    "system": "You are a professional programming assistant.",
    "messages": [
      {"role": "user", "content": "Write a quicksort in Python"}
    ]
  }'

Python

anthropic_chat.py


import anthropic
 
client = anthropic.Anthropic(
    base_url="https://api.hao.ai/anthropic",
    api_key="<your HAOAI_API_KEY>"
)
 
message = client.messages.create(
    model="anthropic/claude-sonnet-4.6",
    max_tokens=1024,
    system="You are a professional programming assistant.",
    messages=[
        {"role": "user", "content": "Write a quicksort in Python"}
    ]
)
 
print(message.content[0].text)

TypeScript

anthropic_chat.ts


import Anthropic from '@anthropic-ai/sdk'
 
const client = new Anthropic({
  baseURL: 'https://api.hao.ai/anthropic',
  apiKey: '<your HAOAI_API_KEY>'
})
 
const message = await client.messages.create({
  model: 'anthropic/claude-sonnet-4.6',
  max_tokens: 1024,
  system: 'You are a professional programming assistant.',
  messages: [
    { role: 'user', content: 'Write a quicksort in Python' }
  ]
})
 
console.log(message.content[0].text)

Response Format


{
  "id": "msg_abc123",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Here's a Python quicksort implementation..."
    }
  ],
  "model": "anthropic/claude-sonnet-4.6",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 25,
    "output_tokens": 200
  }
}

Streaming

Python

anthropic_stream.py


with client.messages.stream(
    model="anthropic/claude-sonnet-4.6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Tell me a story"}]
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

Supported Models

Model	Description
`anthropic/claude-opus-4.6`	Claude Opus 4 — Most capable
`anthropic/claude-sonnet-4.6`	Claude Sonnet 4 — Balanced performance
`anthropic/claude-sonnet-4.5`	Claude Sonnet 4.5 — Supports manual thinking
`anthropic/claude-haiku-4.5`	Claude Haiku 4.5 — Fast responses

HaoAI’s Anthropic protocol supports all native features, including Vision, Tool Use, Prompt Caching, Extended Thinking, and more. For the complete available model list, use the Models API or Model Plaza.