Execute instructions using prompt (OpenAI format)

POST

/v1/chat/completions

Supports inputting instructions in user prompts to execute specific functions, including operations such as executing commands, deploying applications, and changing the LLM model of the sandbox.

Instruction List

Instruction	Function Description
`/commands`	Execute Shell commands
`/deploy`	Deploy application
`/model`	Change the LLM model used by the sandbox
`/max_thinking_token`	Change the maximum thinking token for sandbox settings (0 means turn off thinking)

Instructions for Use

Important Note: The prompt must start with a command keyword, such as /commands, /model, /deploy.

Example

/commands 'npm install express'

/model glm-4.6

Conversation ID (session_id) before /deploy

Billing Instructions

Price: 0.0005 PTC per second of sandbox usage

Request

Header Params

Body Params application/jsonRequired

Examples

基础的cmd指令示例带参数的cmd指令修改沙盒使用的llm模型基础的部署指令部署指令并配置env变量开启claude code思考（值越大效果越好，token花费也越高）关闭claude code思考

{
    "model": "302-sandbox-xxxxxxx",
    "messages": [
        {
            "role": "user",
            "content": "/commands 'ls -la'"
        }
    ]
}

Request Code Samples

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

cURL

curl --location --request POST 'https://api.302.ai/v1/v1/chat/completions' \
--header 'Authorization: Bearer ' \
--header 'session_id;' \
--header 'Content-Type: application/json' \
--data-raw '{
    "model": "302-sandbox-xxxxxxx",
    "messages": [
        {
            "role": "user",
            "content": "/commands '\''ls -la'\''"
        }
    ]
}'

Responses

🟢200Success

application/json

Body

Example

{}

Modified at 2025-12-19 11:27:18

Call Claude-Code (OpenAI format)

Call Claude-Code (Claude format)

Execute instructions using prompt (OpenAI format)

Instruction List#

Instructions for Use#

Example#

Billing Instructions#

Request

Request Code Samples

Responses

Instruction List

Instructions for Use

Example

Billing Instructions