StartAgent-阿里云帮助中心

Start AI Agent

Try it now

Try this API in OpenAPI Explorer, no manual signing needed. Successful calls auto-generate SDK code matching your parameters. Download it with built-in credential security for local usage.

Test

RAM authorization

The table below describes the authorization required to call this API. You can define it in a Resource Access Management (RAM) policy. The table's columns are detailed below:

Action: The actions can be used in the Action element of RAM permission policy statements to grant permissions to perform the operation.
API: The API that you can call to perform the action.
Access level: The predefined level of access granted for each API. Valid values: create, list, get, update, and delete.
Resource type: The type of the resource that supports authorization to perform the action. It indicates if the action supports resource-level permission. The specified resource must be compatible with the action. Otherwise, the policy will be ineffective.
- For APIs with resource-level permissions, required resource types are marked with an asterisk (*). Specify the corresponding Alibaba Cloud Resource Name (ARN) in the Resource element of the policy.
- For APIs without resource-level permissions, it is shown as All Resources. Use an asterisk (*) in the Resource element of the policy.
Condition key: The condition keys defined by the service. The key allows for granular control, applying to either actions alone or actions associated with specific resources. In addition to service-specific condition keys, Alibaba Cloud provides a set of common condition keys applicable across all RAM-supported services.
Dependent action: The dependent actions required to run the action. To complete the action, the RAM user or the RAM role must have the permissions to perform all dependent actions.

Action

Access level

Resource type

Condition key

Dependent action

rtc:StartAgent

create

*App

acs:rtc::{#accountId}:app/{#AppId}

None

Request parameters

Parameter	Type	Required	Description	Example
AppId	string	Yes	The application ID. You can create and query this ID in the console. Only one application ID is supported.	aoe****
ChannelId	string	Yes	The ID of the channel to join. Only one channel ID is supported.	yourChannelId
TaskId	string	Yes	The task ID.	yourTaskId
TemplateId	string	Yes	The template ID.	76dasgb****
RtcConfig	object	Yes	The configuration for the agent to join an RTC channel.	2
UserId	string	Yes	The user ID for the agent in the RTC channel.	423341
TargetUserIds	array	No	A list of user IDs that the agent subscribes to. Currently, only a single user ID is supported.
	string	No	A user ID.	41411
UserInactivityTimeout	integer	No	The maximum duration of user inactivity, in seconds. The agent automatically stops and leaves the channel if either of the following conditions is met for this duration: All users in the `targetUserIds` list have left the channel. No users from the `targetUserIds` list have joined the channel. Default: 30. Range: [5, 180].	30
VoiceChatConfig	object	No	The configuration for AI voice chat.
ChatMode	integer	No	`1`: Natural conversation mode (default). `2`: Walkie-talkie mode.	1
InterruptMode	integer	No	`1`: Enables voice interruption (default). `2`: Disables voice interruption.	1
Greeting	string	No	The agent's initial greeting.	早上好，我的朋友
ASRConfig	object	No	Configuration for Automatic Speech Recognition (ASR).
SourceLanguage	string	No	The language is automatically detected, so this parameter is not required.	zh
SemanticPunctuationEnabled	boolean	No	Enables semantic punctuation for sentence segmentation. If `false`, segmentation is based on Voice Activity Detection (VAD).	false
MaxSentenceSilence	integer	No	The silence duration threshold, in milliseconds, for VAD-based sentence segmentation. Default: 800. Recommended range: [200, 6000].	800
VocabularyId	string	No	The ID of the custom vocabulary to use.	vocab-xxx-24ee19fa8cfb4d52902170a0xxxxxxxx
LanguageHints	array	No	If this parameter is not set, the model automatically detects the language. Supported language codes: • `zh`: Chinese (supports dialects including Shanghainese, Wu, Minnan, Northeastern Mandarin, Gansu, Guizhou, Henan, Hubei, Hunan, Jiangxi, Ningxia, Shanxi, Shaanxi, Shandong, Sichuan, Tianjin, Yunnan, and Cantonese) • `en`: English • `ja`: Japanese • `yue`: Cantonese • `ko`: Korean
	string	No	A language code.	zh
VadConfig	object	No	Configuration for Voice Activity Detection (VAD).
InterruptSpeechDuration	integer	No	The minimum duration, in milliseconds (ms), of continuous human speech required to trigger an interruption. This helps prevent false interruptions. Default: 300. Range: [0, 3000].	300
LLMConfig	object	No	Configuration for the Large Language Model (LLM).
Vendor	string	No	The name of the cloud service provider. Examples: `aliyun` (for Model Studio), `thirdparty`.	aliyun
Model	string	No	The name of the model. For the `aliyun` vendor, supported models include `qwen-plus`, `deepseek-r1`, and `qwen-omni-turbo`.	qwen-plus
ApiKey	string	No	The API key from the cloud service provider.	xxxxxxxxxxx
Temperature	number	No	The sampling temperature, which controls the randomness of the generated text. A higher value increases randomness. Range: [0, 2). Default: 0.7. Note: This parameter is not supported for the `deepseek-r1` model.	0.7
TopP	number	No	Controls output diversity via nucleus sampling. The model samples from the smallest set of tokens whose cumulative probability mass exceeds this value. A higher value results in a more diverse output. Range: (0, 1]. Default: 0.8.	0.8
MaxToken	integer	No	The maximum number of tokens for the generated response. For `deepseek` and `qwen` models, the upper limit is 8192. Default: 8192.	500
HistoryDepth	integer	No	The number of previous conversation turns to include as context for the model. Default: 3.	3
Prompt	string	No	A system prompt that instructs the model on its role, behavior, or persona.	xxxx
Url	string	No	The access URL for a third-party LLM.	https://xxxxx
Params	object	No	A key-value map for custom parameters. This parameter is deprecated; use `BizParams` instead.	{\"verification_code\": \"xxx\"}
BizParams	any	No	Custom parameters.
AppId	string	No	The agent application ID. You can specify either `Model` or `AppId`. If both are provided, `AppId` takes precedence.	e****
ToolExecutionConfig	any	No
Tools	array	No
	any	No
TTSConfig	object	No	Configuration for Text-to-Speech (TTS).
Vendor	string	No	The cloud service provider. Example: `aliyun` (for Model Studio).	aliyun
Model	string	No	The name of the TTS model. Example: `cosyvoice-v1`.	cosyvoice-v1
ApiKey	string	No	The API key from the cloud service provider.	xxxxxx
Voice	string	No	The voice for the synthesized speech. If you use the `cosyvoice-v1` model from Alibaba Cloud, the default voice is `longxiaoxia`. Available voices: `longwan`, `longcheng`, `longhua`, `longxiaochun`, `longxiaoxia`, `longxiaocheng`, `longxiaobai`, `longlaotie`, `longshu`, `longshuo`, `longjing`, `longmiao`, `longyue`, `longyuan`, `longfei`, `longjielidou`, `longtong`, `longxiang`, `loongstella`, `loongbella`.	longxiaoxia
Volume	integer	No	The volume of the synthesized speech. Range: [0, 100]. Default: 50.	50
Rate	number	No	The speech rate. Range: [0.5, 2.0]. Default: 1.0.	1
Pitch	number	No	The pitch of the synthesized speech. Range: [0.5, 2.0]. Default: 1.0.	1
FilterBrackets	array	No	Specifies bracket types whose enclosed content should be excluded from speech synthesis.
	integer	No	An integer representing the bracket type to filter. • `1`: Chinese parentheses（） • `2`: English parentheses () • `3`: Chinese square brackets【】 • `4`: English square brackets [] • `5`: English curly braces {}	[1,2]
AgentSilenceConfig	object	No	Configures how the agent responds to silence, such as when a user is in the channel but not speaking. You can configure the agent to play an audio prompt or send a webhook notification.
Enable	boolean	No	Specifies whether to enable the silence configuration.	false
AlertTimeout	integer	No	The timeout in seconds before triggering a silence alert. The alert can be re-triggered if the silence continues. Range: [5, 600]. Default: 0 (disabled).	0
Strategy	integer	No	The strategy for the silence alert: • `1` (default): Plays the message specified in `Content`. • `2`: Sends the `Content` message and conversation history to the LLM, and then plays the generated response.	1
Content	string	No	The content of the silence alert message.	**
WebhookTriggerTimeout	integer	No	The timeout in seconds before triggering a silence webhook. The webhook can be re-triggered if the silence continues. Range: [10, 600]. Default: 0 (disabled).	0
AmbientSoundConfig	object	No	Configuration for ambient sound.
SoundId	string	No	Specifies the ambient sound to play. If this parameter is not provided, ambient sound is disabled. Built-in sounds: Set to `white_noise` or `office`. For custom sounds, provide a `soundId` that you have configured in the console with a corresponding download URL. (This feature is planned for a future release).	white_noise
Volume	integer	No	The volume of the ambient sound. Range: [0, 100]. Default: 50.	50
InterruptConfig	object	No	Configuration for interruption handling.
SemanticsInterrupt	boolean	No	Specifies whether to enable semantic interruption. Default: `true`.	true
BackChannelConfig	object	No	Configuration for back-channeling.
UserTurnEnd	boolean	No	Specifies whether to enable back-channeling after the user finishes speaking. Default: `false`.	false
Version	integer	No
EnableVideoUnderstanding	boolean	No	Specifies whether to enable video understanding. If enabled, the LLM must support multimodal inputs.
preferVideo	integer	No	When video understanding is enabled, specifies which video stream to prioritize if both camera and screen sharing streams are active. The agent can process only one stream at a time. `1`: Prioritize the camera stream. `2`: Prioritize the screen sharing stream. Default: `1`.

Response elements

Element	Type	Description	Example
	object	The response data.
RequestId	string	The unique ID of the request.	6159ba01-6687-4fb2-a831-f0cd8d188648

Examples

Success response

JSON format

{
  "RequestId": "6159ba01-6687-4fb2-a831-f0cd8d188648"
}

Error codes

See Error Codes for a complete list.

Release notes

See Release Notes for a complete list.