Claude API

This platform provides a chat endpoint compatible with the Anthropic Claude Messages API.

This document is intended for client-side integrators and focuses on the public request path, authentication method, request parameters, and example payloads. Field names and JSON structure are kept as close as possible to the native Claude Messages API.

Overview:

Endpoint: POST /v1/messages
Capability type: chat / text generation
Response modes:
- When stream = false or omitted: standard JSON response
- When stream = true: text/event-stream SSE response
Protocol style: the request body follows the native Claude Messages API structure and does not wrap content in an extra input field

1. API Overview

HTTP method: POST
Request path: /v1/messages
Content-Type: application/json
Response type:
- Non-streaming: application/json
- Streaming: text/event-stream

2. Authentication and Headers

Example headers:

Authorization: Bearer YOUR_REACH_API_KEY
anthropic-version: 2023-06-01
Content-Type: application/json

At least one of Authorization or x-api-key must be provided.

Header details:

Header	Required	Description
`Authorization`	No	Platform API key in the format `Bearer sk-xxxxxx`
`x-api-key`	No	You can also pass the platform API key directly in this header
`anthropic-version`	Recommended	For Claude Messages API client compatibility, `2023-06-01` is recommended
`Content-Type`	Yes	Must be `application/json`

3. Request Body

The request body follows the native Claude Messages API object structure:

{
  "model": "YOUR_MODEL_ID",
  "messages": [
    {
      "role": "user",
      "content": "Please introduce yourself in one sentence."
    }
  ],
  "max_tokens": 1024,
  "temperature": 0.2,
  "stream": false
}

3.1 Top-Level Parameters

Field	Type	Required	Description
`model`	string	Yes	Model ID
`messages`	`array<object>`	Yes	Chat message list
`system`	`string` or `array<object>`	No	System prompt
`max_tokens`	integer	Yes	Maximum number of output tokens, minimum value `1`
`temperature`	number	No	Sampling temperature
`top_p`	number	No	Nucleus sampling parameter
`top_k`	integer	No	Top-k sampling parameter
`stream`	boolean	No	Whether to return SSE streaming output
`stop_sequences`	`array<string>`	No	Additional stop sequences
`tools`	`array<object>`	No	Tool definitions
`tool_choice`	object	No	Tool calling strategy
`thinking`	object	No	Reasoning budget or thinking-mode configuration
`metadata`	object	No	Invocation metadata

3.2 `messages` Parameter

Each message uses the following structure:

Field	Type	Required	Description
`role`	string	Yes	Role. Common values: `user`, `assistant`
`content`	`string` or `array<object>`	Yes	Message content. Can be plain text or an array of content blocks

Text example:

{
  "role": "user",
  "content": "Write a short product introduction within 50 words."
}

Content block example:

{
  "role": "user",
  "content": [
    {
      "type": "text",
      "text": "Please describe this image."
    }
  ]
}

Common content block types:

`type`	Typical fields	Description
`text`	`text`	Text content
`image`	`source`	Image input
`tool_use`	`id`, `name`, `input`	Tool call initiated by the model
`tool_result`	`tool_use_id`, `content`	Tool execution result returned to the model

4. Request Examples

4.1 Standard Non-Streaming Request

curl -X POST "https://direct.reachapi.ai/v1/messages" \
  -H "Authorization: Bearer YOUR_REACH_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_ID",
    "messages": [
      {
        "role": "user",
        "content": "Introduce ReachAPI in one sentence."
      }
    ],
    "max_tokens": 512,
    "temperature": 0.2
  }'

4.2 Streaming Request

curl -X POST "https://direct.reachapi.ai/v1/messages" \
  -H "Authorization: Bearer YOUR_REACH_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_ID",
    "messages": [
      {
        "role": "user",
        "content": "Summarize today'\''s priorities in three lines."
      }
    ],
    "max_tokens": 512,
    "stream": true
  }'

4.3 Request with Tool Definitions

curl -X POST "https://direct.reachapi.ai/v1/messages" \
  -H "Authorization: Bearer YOUR_REACH_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_ID",
    "messages": [
      {
        "role": "user",
        "content": "Help me check the weather in Shanghai."
      }
    ],
    "max_tokens": 512,
    "tools": [
      {
        "name": "get_weather",
        "description": "Get weather information for a city",
        "input_schema": {
          "type": "object",
          "properties": {
            "city": {
              "type": "string"
            }
          },
          "required": ["city"]
        }
      }
    ],
    "tool_choice": {
      "type": "auto"
    }
  }'

5. Non-Streaming Response

On success, the API returns a standard message object:

{
  "id": "msg_01ABCDEF",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "ReachAPI is a unified AI gateway platform."
    }
  ],
  "model": "YOUR_MODEL_ID",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 32,
    "output_tokens": 18,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 0
  }
}

6. Streaming Response

When stream = true, the endpoint returns text/event-stream. Typical events look like this:

event: message_start
data: {"type":"message_start","message":{"id":"msg_1","type":"message","role":"assistant","content":[],"model":"YOUR_MODEL_ID","stop_reason":null,"usage":{"input_tokens":7,"output_tokens":0}}}

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

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"end_turn"},"usage":{"output_tokens":13}}

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

7. Error Response

When a request fails, the API returns a JSON error object. Example:

{
  "type": "error",
  "error": {
    "type": "invalid_request_error",
    "message": "Field `model` is required"
  }
}

Common scenarios:

Missing or invalid API key
Request body is not valid JSON
Missing model
Missing messages

8. Compatibility Notes

This document lists commonly used fields and example payloads. Actual field names and structures follow the Claude Messages API.
If you use advanced features such as tool calling or thinking budgets, confirm that the target model actually supports them.
If your client is built on top of the Anthropic or Claude SDK, passing anthropic-version is recommended.

9. Supported Models and Reference Pricing

The following prices are summarized from the unified pricing document and apply to the Claude text models currently listed. Pricing was checked on 2026-04-17. For detailed pricing rules, refer to claude.md.

Model	Input	Cache write 5m	Cache read	Output
`claude-haiku-4-5`	`$1 / 1M tokens`	`$1.25 / 1M tokens`	`$0.10 / 1M tokens`	`$5 / 1M tokens`
`claude-sonnet-4-5`	`$3 / 1M tokens`	`$3.75 / 1M tokens`	`$0.30 / 1M tokens`	`$15 / 1M tokens`
`claude-sonnet-4-6`	`$3 / 1M tokens`	`$3.75 / 1M tokens`	`$0.30 / 1M tokens`	`$15 / 1M tokens`
`claude-opus-4-5-20251101`	`$5 / 1M tokens`	`$6.25 / 1M tokens`	`$0.50 / 1M tokens`	`$25 / 1M tokens`
`claude-opus-4-6`	`$5 / 1M tokens`	`$6.25 / 1M tokens`	`$0.50 / 1M tokens`	`$25 / 1M tokens`

Notes:

Cache write 5m refers to the prompt cache write price for the 5-minute tier
claude-opus-4-5-20251101 is a snapshot model ID and can be used for version-pinned requests
For longer cache durations or extended context pricing, refer to the detailed pricing document

10. References

Anthropic Messages API: https://docs.anthropic.com/en/api/messages
Anthropic Messages examples: https://docs.anthropic.com/en/api/messages-examples
Unified platform pricing document: claude.md

Claude API

On this page