Anthropic Messages

Send messages and Content Blocks with the Anthropic Messages-compatible structure.

The Messages API creates model responses from user and assistant messages and Content Blocks. It supports text, images, tool calls, streaming output, and extended thinking. Available content blocks, tools, and advanced parameters depend on the selected model and the models visible to the current API key.

POST/v1/messages

Generate the next assistant response from user and assistant messages. Message content can be a string or an array of text, image, tool, and document Content Blocks.

The API supports standard JSON responses and SSE streaming events. Image input, tool calling, and extended thinking depend on the selected model.

Authorization

anthropicApiKey

x-api-key<token>

Pass the API Key in the x-api-key Header.

In: header

Header Parameters

anthropic-version*string

Specify the API version to use.

Default"2023-06-01"

anthropic-beta?string

Specify the beta version to use. You can use commas to separate multiple values, or send the header multiple times.

Request Body

Messages represents text, images, tools, and extended thinking with Content Blocks. Choose a view for your task.

application/json

Text chat

Use user and assistant messages for single-turn or multi-turn conversations; provide system instructions in the top-level field.

model: A Messages-capable model ID visible to the current API key.
max_tokens: The maximum number of tokens the response may generate.
messages[]: User and assistant messages in chronological order.

application/json

{
  "model": "MODEL_ID",
  "max_tokens": 1024,
  "messages": [
    {
      "role": "user",
      "content": "Hello! Please introduce yourself."
    }
  ]
}

Basic text chat. messages can contain strings or text Content Blocks.

model*string

The model ID to use. Select a Messages-capable model visible to the current API key.

Length1 <= length <= 256

max_tokens*integer

Maximum number of tokens allowed to be generated. Different models have different limitations; Set to 0 can only fill the prompt cache without generating a response.

Range0 <= value

messages*array<>

Text conversation messages in chronological order. content can be a string or an array of text Content Blocks.

Items1 <= items

metadata?

Metadata associated with the request.

cache_control?|null

Top-level cache control, which is automatically applied to the last cacheable content block in the request. Availability depends on the model and channel.

container?||null

Cross-request reuse containers, or specify Skills to load into the container. Availability depends on the model, channel, and account.

inference_geo?string|null

Specify the geographic region for inference; Use the default area of the account when not set. Whether available depends on the account and channel.

output_config?

Model output configuration. Field availability depends on the selected model and channel.

service_tier?string

The service level used by the request. Auto can use priority capacity when available, standard_only only uses standard capacity; The actual support scope depends on the account and channel.

Value in

"auto"
"standard_only"

stop_sequences?array<string>

Custom stop generating text sequences.

stream?boolean

Whether to use the server to send events and return responses step by step.

Defaultfalse

system?|array<>

Top-level system prompt. There is no system role in the Messages list.

temperature?number

Deprecated

has been deprecated. Models released after Claude Opus 4.6 do not support custom temperature; To accommodate old requests, accept 1.0, and other values may return 400. It is recommended that new models omit this field.

Range0 <= value <= 1

thinking?||

Extended thinking configuration. New models prioritize adaptive; Older models may continue to use enabled and budget_tokens.

tool_choice?||

tools?array<|||||||>

Custom tools or built-in tools that models can use.

top_k?integer

Deprecated

has been deprecated. Models released after Claude Opus 4.6 may reject any top_k value and return 400. It is recommended that new models omit this field.

Range1 <= value

top_p?number

Deprecated

Deprecated. Models released after Claude Opus 4.6 only retain limited compatibility behavior for settings close to the default value; other values may return 400. It is recommended that new models omit this field.

Range0 < value < 1

Image, tool, and extended-thinking support can vary by model and provider channel. Start with a minimal text request, then enable advanced fields one at a time.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "MODEL_ID",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Hello! Please introduce yourself."      }    ]  }'

{
  "content": [
    {
      "type": "text",
      "text": "Hello! I am an AI assistant."
    }
  ],
  "id": "msg_example",
  "model": "MODEL_ID",
  "role": "assistant",
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "type": "message",
  "usage": {
    "input_tokens": 12,
    "output_tokens": 16,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 0
  }
}

Anthropic Messages documentationReview the upstream protocol and recent field changes.

Authorization

Header Parameters

Request Body

Response Body

200application/json

400application/json

401application/json

429application/json

500application/json

`application/json`

`application/json`

`application/json`

`application/json`