AgentKit API
火山引擎Call the Volcano Engine AgentKit agent API and get a conversational agent with tool calling and memory running in 30 minutes.
Build agents with low-code/0 no-code and publish them with one click to channels like WeChat Official Accounts to provide customer service Q&A.
Coze Programming is an agent development platform for low-code scenarios. By arranging the agent's persona and reply logic on the orchestration page, selecting a model, adding plugins, binding a knowledge base, and debugging, users can build a customer-service agent without writing any code. According to the official tutorial, the finished agent can be published as a WeChat Official Account (Subscription Account) customer service bot that responds to subscriber messages in real time, supporting product Q&A and visual question answering based on image understanding. It is suitable for operators who need to quickly launch an intelligent customer service bot on their Official Account; before publishing, you need a WeChat Official Account in normal status and a Developer ID (AppID).
Agents in Coze Programming are 'applications built on model-based technology.' A default model is preset by the system during setup, and it can also be flexibly switched according to model capabilities to respond to user questions more accurately. Plugins extend agent capabilities, enabling them to perform tasks such as search, file processing, and schedule management; the tutorial scenario uses a Search plugin (performing an online search to generate replies when the knowledge base has no match) and an Image Understanding plugin (adding image recognition to models without built-in visual understanding). Two types of knowledge bases are supported: a document knowledge base automatically collected via URL (the example uploads documents from the Coze Programming Documentation Center), and a spreadsheet-format FAQ knowledge base. The example customer service agent in the tutorial has two capabilities: product Q&A (answering usage questions and troubleshooting issues) and visual understanding (recognizing images such as error screenshots sent by users and providing solutions). Applicable scenarios include product inquiries and after-sales support in e-commerce, and course consultations and material lookups in education. Note that this article is compiled from the official tutorial documentation and only covers the publishing flow for the WeChat Official Account (Subscription Account) channel; other publishing channels and billing information are not covered on this page.
Content creators, media organizations, e-commerce and education operators who need to build intelligent customer service for WeChat Official Accounts, as well as users with no programming background who want to build agents.
Log in to Coze Programming and select a workspace, then create a low-code agent via 'New Project - Agent Development'.
Set the agent's persona and reply logic by writing manually, generating with AI, or applying a prompt template.
Select or switch models based on model capabilities and business scenarios.
Add the Search plugin (online search when the knowledge base has no match) and the Image Understanding plugin (recognizing images sent by users).
Create and bind a document knowledge base and a spreadsheet knowledge base, for example by automatically collecting and uploading product documents and FAQs via URL.
Set the opening message, then chat with the agent in the debugging area to verify its Q&A performance.
Click Publish at the top right of the orchestration page, select the 'WeChat Official Account (Subscription Account)' channel, enter the AppID, and save.
Use the Official Account administrator's WeChat to scan the QR code and authorize the platform account, then confirm publishing.
On the publishing page, click Chat Now and scan the QR code to chat with the Official Account customer service agent and try it out.
请核对官方的地区支持、数据政策和商用约定。
费用与套餐:以官网当前说明为准,本站不提供套餐价格保证。