# Alibaba Cloud Qwen Intelligence API

> Alibaba Cloud Qwen Intelligence API on Apizie: 4 endpoints (Calls a published Creative Agent with support for Lite and Dedicated editions, and returns execution…, Calls a published Mobile-Use Agent instance, which supports the Standard Edition and Dedicated Edition.…, Invokes a published Mobile Planner Agent using OpenAI-Compatible Chat Completions messages to initiate a…, version) returned as clean JSON through one API key. Send your own provider key in the x-upstream-key header. Call it over REST at https://zieapi.vercel.app/v1/cloud:aliyun-maasqiservice/{endpoint} or from Claude, ChatGPT and Cursor through the Apizie MCP server. Free plan: 1,000 requests a month.

- Project key: `cloud:aliyun-maasqiservice`
- Category: Cloud
- Status: byok
- Source: https://api.alibabacloud.com/product/maasqiservice
- Page: https://zieapi.vercel.app/api/cloud/aliyun-maasqiservice
- OpenAPI: https://zieapi.vercel.app/api/cloud/aliyun-maasqiservice/openapi.json

Alibaba Cloud OpenAPI for Qwen Intelligence (maasqiservice, version 2026-08-31). Calls are signed with ACS3-HMAC-SHA256 using your AccessKey ID and secret.

## Endpoints

### Calls a published Creative Agent with support for Lite and Dedicated editions, and returns execution…
`POST https://zieapi.vercel.app/v1/cloud:aliyun-maasqiservice/aigc-chat-completion-stream` (byok)

| Parameter | Type | Required | Description |
|---|---|---|---|
| endpoint | string | no | Regional endpoint, for example https://maasqiservice.cn-beijing.aliyuncs.com |
| X-QI-Agent-Api-Key | string | yes | The instance access key. The key must belong to the Creative Agent instance being called. This key is different from the AccessKey Secret… |
| X-QI-Instance-Id | string | yes | The published Creative Agent instance ID. Must correspond to the X-QI-Agent-Api-Key |
| X-QI-Session-Id | string | no | The session identifier. Can be omitted for the first turn. Obtain it from the qi_session_id field in the SSE response and pass it back in… |
| messages | array | yes | The conversation message array. Creative Agent uses the last user message as the current objective and executes it in combination with the… |
| metadata | object | no | The extended input for this image creation request. Use the parameters object to set generation parameters. Do not use this field to pass… |
| model | string | no | A compatibility placeholder field. Can be omitted or set to an empty string. You cannot use this field to switch the model used by the… |
| stream | string | yes | Specifies whether to use SSE streaming responses. This parameter must be set to true for this operation |
| streamOptions | object | no | The HTTP JSON key is stream_options. Set include_usage to retrieve the token usage for this call |

### Calls a published Mobile-Use Agent instance, which supports the Standard Edition and Dedicated Edition.…
`POST https://zieapi.vercel.app/v1/cloud:aliyun-maasqiservice/gui-chat-completion-stream` (byok)

| Parameter | Type | Required | Description |
|---|---|---|---|
| endpoint | string | no | Regional endpoint, for example https://maasqiservice.cn-beijing.aliyuncs.com |
| X-QI-Agent-Api-Key | string | yes | The access key of the agent instance, which is used to authenticate QICloud Customer API operations |
| X-QI-Instance-Id | string | yes | The ID of the published Mobile-Use Agent instance. This ID must correspond to the X-QI-Agent-Api-Key |
| X-QI-Session-Id | string | no | The optional session identifier. You can omit this parameter in the first step, and it is generated by the server. You must pass back the… |
| allowedTokenIds | array | no | The whitelist of token IDs that are allowed to be generated |
| badWords | array | no | The list of words or strings that are banned from being generated |
| chatTemplateKwargs | object | no | The vLLM or OpenAI-compatible template parameter object. The request values override the default values with the same names in the model… |
| frequencyPenalty | number | no | The frequency penalty parameter, which is passed through to the primary model of the Mobile-Use Agent |
| ignoreEos | string | no | Specifies whether to ignore the end-of-sequence (EOS) token and continue generation |
| includeReasoning | string | no | Specifies whether to request the upstream to generate the reasoning field. The client-facing results of the Mobile-Use Agent do not… |
| logprobs | string | no | The model compatibility parameter. The JSON field is logprobs. Whether this parameter is accepted and takes effect depends on the model… |
| maxCompletionTokens | integer | no | The OpenAI-compatible maximum number of completion tokens. If provided, the value must be greater than 0 |
| maxTokens | integer | no | The OpenAI-compatible maximum number of generated tokens. If provided, the value must be greater than 0 |
| messages | array | yes | The messages for the current Mobile-Use Agent step. The first step can only contain one user message. Subsequent steps can only contain… |
| metadata | object | no | The metadata for the Mobile-Use Agent step. Supports screen_width, screen_height, available_apps, and the compatibility field… |
| minP | number | no | The vLLM Min-P sampling parameter |
| minTokens | integer | no | The minimum number of tokens to generate |
| mmProcessorKwargs | object | no | The multimodal preprocessor parameters, passed through to the Mobile-Use Agent primary model as-is by key |
| model | string | no | Reserved for OpenAI client compatibility. This parameter is ignored during routing. The model is determined by the published configuration… |
| n | integer | no | Set this parameter to 1. Whether this parameter is accepted depends on the model. The current Mobile-Use Agent response does not return… |
| parallelToolCalls | string | no | The model-compatible switch to invoke parallel tool calling. This parameter does not expand the set of Mobile-Use Agent actions active for… |
| presencePenalty | number | no | The presence penalty parameter, passed through to the Mobile-Use Agent primary model |
| promptLogprobs | integer | no | A model-compatible parameter. The JSON field is prompt_logprobs. Whether this parameter is accepted depends on the model. The current… |
| reasoningEffort | string | no | The OpenAI-compatible reasoning effort, such as low, medium, or high |
| repetitionPenalty | number | no | The vLLM repetition penalty parameter |
| responseFormat | object | no | The model-compatible structured output options. This parameter does not change the final Mobile-Use Agent tool calling structure when you… |
| seed | integer | no | The random seed |
| skipSpecialTokens | string | no | Specifies whether to skip special tokens |
| stop | array | no | The stop sequences. This parameter is compatible with a single string or an array of strings. The array format is shown here |
| stopTokenIds | array | no | The list of vLLM stop token IDs |
| stream | string | yes | Specifies whether to use Server-Sent Events (SSE) for streaming responses. You must set this parameter to true for this operation |
| streamOptions | object | no | The OpenAI-compatible streaming options |
| structuredOutputs | object | no | The vLLM structured output constraints |
| temperature | number | no | The sampling temperature, which overrides the default value of the Mobile-Use Agent primary model |
| topK | integer | no | The vLLM Top-K sampling parameter |
| topLogprobs | integer | no | A model-compatible parameter. The JSON field is top_logprobs. Whether this parameter is accepted depends on the model. The current… |
| topP | number | no | The Top-P nucleus sampling parameter, which overrides the default value of the Mobile-Use Agent primary model |

### Invokes a published Mobile Planner Agent using OpenAI-Compatible Chat Completions messages to initiate a…
`POST https://zieapi.vercel.app/v1/cloud:aliyun-maasqiservice/pa-chat-completion-stream` (byok)

| Parameter | Type | Required | Description |
|---|---|---|---|
| endpoint | string | no | Regional endpoint, for example https://maasqiservice.cn-beijing.aliyuncs.com |
| X-QI-Agent-Api-Key | string | yes | The Agent instance access key used for QICloud Customer API authentication |
| X-QI-Instance-Id | string | yes | The published Mobile Planner Agent instance ID. Must correspond to the X-QI-Agent-Api-Key |
| X-QI-Session-Id | string | no | The optional session identifier. Can be omitted for the first turn, in which case the server generates one. Pass it back in subsequent… |
| allowedTokenIds | array | no | The whitelist of token IDs allowed for generation |
| badWords | array | no | The list of words or strings prohibited from generation |
| chatTemplateKwargs | object | no | The vLLM/OpenAI-Compatible template parameter object. Request values are merged with model defaults, with request keys taking precedence,… |
| frequencyPenalty | number | no | The frequency penalty parameter passed through to the Mobile Planner Agent model |
| ignoreEos | string | no | Specifies whether to ignore EOS and continue generation |
| includeReasoning | string | no | Specifies whether to request reasoning content from the upstream. The actual capability depends on the model |
| logprobs | string | no | A model-compatible parameter with the JSON field name logprobs. Whether it is accepted and takes effect depends on the model used by the… |
| maxCompletionTokens | integer | no | The maximum number of completion tokens. Must be greater than 0 if provided, and is passed as an OpenAI-compatible model parameter |
| maxTokens | integer | no | The OpenAI-compatible maximum number of generated tokens. Must be greater than 0 if provided |
| messages | array | yes | The list of conversation messages. The first call can submit one turn or an existing multi-turn conversation. When continuing with an… |
| minP | number | no | The vLLM Min-P sampling parameter |
| minTokens | integer | no | The minimum number of tokens to generate |
| mmProcessorKwargs | object | no | The multimodal preprocessor parameters passed through to vLLM as-is by key |
| model | string | no | Reserved for OpenAI client compatibility. Ignored during routing. The actual model is determined by the published instance configuration.… |
| n | integer | no | Set this parameter to 1. Whether the model accepts this parameter depends on the model. The current Mobile Planner Agent response does not… |
| parallelToolCalls | string | no | Specifies whether the model is allowed to generate multiple tool calling requests in parallel. The model invokes tools concurrently when… |
| presencePenalty | number | no | The presence penalty parameter, passed through to the Mobile Planner Agent model |
| promptLogprobs | integer | no | A model compatibility parameter. The JSON field is prompt_logprobs. Whether the model accepts this parameter depends on the model. The… |
| reasoningEffort | string | no | The OpenAI-compatible reasoning effort level, such as low, medium, or high |
| repetitionPenalty | number | no | The vLLM repetition penalty parameter |
| responseFormat | object | no | The structured response format, such as json_object or json_schema |
| seed | integer | no | The random seed |
| skipSpecialTokens | string | no | Specifies whether to skip special tokens |
| stop | array | no | The stop sequences. Compatible with a single string or an array of strings. This field is displayed in array format |
| stopTokenIds | array | no | The vLLM stop token ID list |
| stream | string | yes | Specifies whether to use SSE streaming responses. This parameter must be set to true for this operation |
| streamOptions | object | no | The OpenAI-compatible streaming options |
| structuredOutputs | object | no | The vLLM structured output constraints |
| temperature | number | no | The sampling temperature. Passed as an OpenAI-compatible model parameter to override the model default value |
| toolChoice | string | no | The tool choice strategy. Valid values: auto, none, required, or a named selection supported by the model |
| tools | array | no | The array of function tool definitions available for this request. Mobile Planner Agent registers each named function in the runtime… |
| topK | integer | no | The vLLM Top-K sampling parameter |
| topLogprobs | integer | no | A model compatibility parameter. The JSON field is top_logprobs. Used for models that support this feature. The current Agent response… |
| topP | number | no | Top-P sampling parameter. Passed as an OpenAI-compatible model parameter to override the model default value |

### version
`POST https://zieapi.vercel.app/v1/cloud:aliyun-maasqiservice/version` (byok)

| Parameter | Type | Required | Description |
|---|---|---|---|
| endpoint | string | no | Regional endpoint, for example https://maasqiservice.cn-beijing.aliyuncs.com |

## Example
```
curl "https://zieapi.vercel.app/v1/cloud:aliyun-maasqiservice/aigc-chat-completion-stream" \
  -H "x-api-key: $APIZIE_KEY" \
  -H "x-upstream-key: $PROVIDER_KEY"
```

## FAQ

### How do I call the Alibaba Cloud Qwen Intelligence API?
Create a free Apizie key, then send POST https://zieapi.vercel.app/v1/cloud:aliyun-maasqiservice/aigc-chat-completion-stream with the header x-api-key. Every endpoint returns { project, data, timestamp }.

### Is the Alibaba Cloud Qwen Intelligence API free on Apizie?
Yes. The Free plan includes 1,000 requests a month at 100 requests a minute, no credit card. Alibaba Cloud Qwen Intelligence itself also needs your own key from aliyuncs.com.

### Can AI agents use the Alibaba Cloud Qwen Intelligence API?
Yes. Add the Apizie MCP server (https://zieapi.vercel.app/mcp) to Claude, ChatGPT, Cursor or VS Code, then ask in plain language. The agent calls run_scraper with project key cloud:aliyun-maasqiservice.

### What is the status of the Alibaba Cloud Qwen Intelligence API?
Needs your own provider key (bring your own key)