原生 Claude API

Anthropic Claude Messages API 格式的请求，使用 JSON 请求体创建消息。请求必须包含 anthropic-version，并使用 Authorization Bearer Token 或 x-api-key 其中一种方式认证；可用字段和具体能力取决于目标模型及上游渠道.

请求参数

请求头

下表列出网关当前支持的请求字段；字段是否生效仍取决于目标模型和上游渠道。

请求体

消息与内容块

messages 中每项的 role 为 user 或 assistant。content 可以是字符串，也可以是内容块数组。

文本消息示例：

{
  "role": "user",
  "content": "介绍一下 prompt caching。"
}

多内容块消息示例：

{
  "role": "user",
  "content": [
    { "type": "text", "text": "描述这张图片。" },
    {
      "type": "image",
      "source": {
        "type": "url",
        "url": "https://example.com/image.png"
      }
    }
  ]
}

Base64 图像将 source.type 设为 base64，并同时提供 media_type 与 Base64 编码的 data：

{
  "type": "image",
  "source": {
    "type": "base64",
    "media_type": "image/png",
    "data": "iVBORw0KGgoAAAANSUhEUg..."
  }
}

工具调用

工具定义至少包含 name 和 input_schema，可选 description。input_schema 使用 JSON Schema 描述工具参数。

{
  "tools": [
    {
      "name": "get_weather",
      "description": "查询指定城市的天气",
      "input_schema": {
        "type": "object",
        "properties": {
          "city": { "type": "string", "description": "城市名称" }
        },
        "required": ["city"]
      }
    }
  ],
  "tool_choice": { "type": "auto" }
}

Claude 返回 stop_reason: "tool_use" 时，遍历 content 找到 tool_use 块，在客户端执行对应工具。下一轮请求应保留完整的 assistant 内容，并追加 tool_result：

{
  "role": "user",
  "content": [
    {
      "type": "tool_result",
      "tool_use_id": "toolu_01ABC",
      "content": "晴，25°C"
    }
  ]
}

tool_choice.type 可设为：

• auto：由 Claude 决定是否调用工具。
• any：要求 Claude 调用任一工具。
• tool：通过 name 指定工具。

扩展思考

根据目标模型支持情况，可以使用 thinking 配置扩展思考：

{
  "thinking": {
    "type": "enabled",
    "budget_tokens": 4096
  }
}

关闭扩展思考：

{
  "thinking": {
    "type": "disabled"
  }
}

budget_tokens 必须小于 max_tokens，并且扩展思考需要目标模型支持。启用扩展思考后，不要同时发送模型不支持的采样参数。

请求体示例

{
  "model": "gpt-5.6-sol",
  "system": "你是一个简洁、准确的助手。",
  "messages": [
    { "role": "user", "content": "用三句话介绍 Claude Messages API。" }
  ],
  "max_tokens": 256,
  "temperature": 0.7,
  "stream": false
}

请求示例代码

curl -X POST "https://10000router.com/v1/messages" \
  -H "Authorization: Bearer $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "gpt-5.6-sol",
    "max_tokens": 256,
    "messages": [
      {"role": "user", "content": "用三句话介绍 Claude Messages API。"}
    ]
  }'

const payload = {
  model: "gpt-5.6-sol",
  max_tokens: 256,
  messages: [
    { role: "user", content: "用三句话介绍 Claude Messages API。" }
  ]
};

const response = await fetch("https://10000router.com/v1/messages", {
  method: "POST",
  headers: {
    Authorization: "Bearer " + process.env.ANTHROPIC_API_KEY,
    "anthropic-version": "2023-06-01",
  },
  body: JSON.stringify(payload)
});

if (!response.ok) throw new Error(await response.text());
console.log(await response.json());

payload := strings.NewReader(`{
  "model": "gpt-5.6-sol",
  "max_tokens": 256,
  "messages": [
    {"role": "user", "content": "用三句话介绍 Claude Messages API。"}
  ]
}`)
req, err := http.NewRequest("POST", "https://10000router.com/v1/messages", payload)
if err != nil {
  log.Fatal(err)
}
req.Header.Set("Authorization", "Bearer "+os.Getenv("ANTHROPIC_API_KEY"))
req.Header.Set("anthropic-version", "2023-06-01")
res, err := http.DefaultClient.Do(req)
if err != nil {
  log.Fatal(err)
}
defer res.Body.Close()
if res.StatusCode >= 400 {
  log.Fatal(res.Status)
}
io.Copy(os.Stdout, res.Body)

import os
import requests

payload = {
    "model": "gpt-5.6-sol",
    "max_tokens": 256,
    "messages": [
        {"role": "user", "content": "用三句话介绍 Claude Messages API。"}
    ],
}

response = requests.post(
    "https://10000router.com/v1/messages",
    headers={
        "Authorization": "Bearer " + os.environ["ANTHROPIC_API_KEY"],
        "anthropic-version": "2023-06-01",
    },
    json=payload,
)
response.raise_for_status()
print(response.json())

var client = java.net.http.HttpClient.newHttpClient();
var payload = "{"
    + "\"model\":\"gpt-5.6-sol\","
    + "\"max_tokens\":256,"
    + "\"messages\":[{\"role\":\"user\","
    + "\"content\":\"用三句话介绍 Claude Messages API。\"}]}";
var request = java.net.http.HttpRequest.newBuilder()
    .uri(java.net.URI.create("https://10000router.com/v1/messages"))
    .header("Authorization", "Bearer " + System.getenv("ANTHROPIC_API_KEY"))
    .header("anthropic-version", "2023-06-01")
    .POST(java.net.http.HttpRequest.BodyPublishers.ofString(payload))
    .build();
var response = client.send(
    request,
    java.net.http.HttpResponse.BodyHandlers.ofString()
);
if (response.statusCode() >= 400) {
    throw new IllegalStateException(response.body());
}
System.out.println(response.body());

using System.Net.Http.Json;

using var client = new HttpClient();
client.DefaultRequestHeaders.TryAddWithoutValidation(
    "Authorization",
    "Bearer " + Environment.GetEnvironmentVariable("ANTHROPIC_API_KEY")
);
client.DefaultRequestHeaders.Add("anthropic-version", "2023-06-01");

var payload = new
{
    model = "gpt-5.6-sol",
    max_tokens = 256,
    messages = new[]
    {
        new { role = "user", content = "用三句话介绍 Claude Messages API。" }
    }
};

var response = await client.PostAsJsonAsync(
    "https://10000router.com/v1/messages",
    payload
);
response.EnsureSuccessStatusCode();
Console.WriteLine(await response.Content.ReadAsStringAsync());

返回响应

响应示例

{
  "id": "msg_01ABC123",
  "type": "message",
  "role": "assistant",
  "model": "gpt-5.6-sol",
  "content": [
    {
      "type": "text",
      "text": "Claude Messages API 用统一消息结构完成多轮对话。"
    }
  ],
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 18,
    "output_tokens": 16
  }
}

{
  "type": "error",
  "error": {
    "type": "invalid_request_error",
    "message": "messages: Field required"
  }
}

{
  "type": "error",
  "error": {
    "type": "rate_limit_error",
    "message": "Request rate limit exceeded"
  }
}

非流式响应（200）

成功响应字段如下：

{
  "id": "msg_01ABC123",
  "type": "message",
  "role": "assistant",
  "model": "gpt-5.6-sol",
  "content": [
    {
      "type": "text",
      "text": "Claude Messages API 用统一消息结构完成多轮对话。"
    }
  ],
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 18,
    "output_tokens": 16
  }
}

响应字段

流式响应

将 stream 设为 true 后，响应类型为 text/event-stream。Claude 使用多个 SSE 事件传递一条消息，客户端应按事件顺序拼接 content_block_delta 中的文本增量。

event: message_start
data: {"type":"message_start","message":{"id":"msg_01ABC123","type":"message","role":"assistant","content":[],"model":"gpt-5.6-sol","stop_reason":null}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Claude Messages"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" API"}}

event: content_block_stop
data: {"type":"content_block_stop","index":0}

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"end_turn"},"usage":{"output_tokens":16}}

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

• message_start：消息开始，包含消息 ID 和模型信息。
• content_block_delta：内容增量。文本增量位于 delta.text。
• message_delta：消息结束前的状态和用量更新。
• message_stop：事件流结束。

工具调用、扩展思考和上游能力可能产生其他事件。客户端应根据 event 或 data.type 分支处理，并安全忽略无法识别的事件；不要把完整 JSON 事件直接拼接成文本。

错误响应

常见状态码包括 400（请求参数错误）、401（认证失败）、429（请求频率限制）和 5xx（上游服务错误）。错误示例见上方状态码标签，响应保持 Anthropic 错误对象结构。

使用 x-api-key 时缺少 anthropic-version，或请求体缺少 model、messages、max_tokens，通常会收到 400。模型不可用、参数不兼容时，也应优先检查模型列表和目标模型能力。