Call Claude-Code (Claude format)

POST

/v1/messages

Using Claude Code through the Anthropic language model interface format only supports stream responses.

Session management

Support custom session ID (session_id)
When the request does not pass session_id, it is considered a new conversation, and the system will automatically assign session_id and return it in the response header

Response description

Claude Code is a multi-round dialogue of Agent type. Compared with the native stream response, it will contain additional types of data such as'user 'and'result' (final task result), which will be encapsulated into text blocks and returned.

Task Result Identification

When the content of a text block begins with the following identifier, it indicates that this is a task result:

\n\n---\n**[Conversation Result]**\n\n

The task result will include the working directory path where the code is stored.

Price: 0.0005 PTC * Number of seconds of sandbox usage + Token fee for LLM model

Request

Header Params

Body Params application/jsonRequired

Example

{
    "model": "302-sandbox-xxxx",
    "max_tokens": 1024,
    "messages": [
        {
            "role": "user",
            "content": "Hello, world"
        }
    ],
    "session_id": "session_1"
}

Request Code Samples

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

curl --location --request POST 'https://api.302.ai/v1/v1/messages' \
--header 'Authorization: Bearer ' \
--header 'session_id;' \
--header 'Content-Type: application/json' \
--data-raw '{
    "model": "302-sandbox-xxxx",
    "max_tokens": 1024,
    "messages": [
        {
            "role": "user",
            "content": "Hello, world"
        }
    ],
    "session_id": "session_1"
}'

Responses

🟢200Success

application/json

Body

Example

{}

Modified at 2025-12-19 11:30:32

Execute instructions using prompt (OpenAI format)

Execute instructions using prompt words (Claude format)

Call Claude-Code (Claude format)

Task Result Identification#

Request

Request Code Samples

Responses

Task Result Identification