Tencent Yuanqi

What is Tencent Yuanqi?

Tencent Yuanqi is a platform launched by Tencent for creating and distributing zero-code AI agents, relying on Tencent Hunyuan as well as other models provided by the platform. Individuals and organizations can use prompts, knowledge bases, plugins, and workflows to develop agents that are capable of engaging in conversations and carrying out tasks.

The new version of Yuanqi focuses on connecting with platforms such as official accounts, WeChat customer service, WeCom, app stores, and APIs. It is more suitable for creators, operators, and corporate innovation teams who wish to quickly test the AI services available within Tencent’s ecosystem.

Main functions of Tencent Yuanqi

• Dialogual intelligent agents:Set roles, goals, response rules, welcome messages, and sample questions.
• WeChat official account AI agent:Authorize the official account and synchronize past articles to provide round-the-clock content Q&A and customer service.
• Knowledge base:Import documents, web pages, Q&A pairs, and database data to enhance responses through retrieval.
• Visual workflow:Intent recognition, knowledge retrieval, large models, conditions, loops, batch processing, and response nodes.
• Plugins and tools:By utilizing the Plugin Plaza capabilities, it is also possible to create custom plugins in accordance with the interface specifications.
• MCP plugin:Connect to external tools and services that comply with the Model Context Protocol.
• Multi-Agent：Create multiple intelligent agents with clear roles, with the main agent responsible for selecting and coordinating the sub-agents.
• Online search:Invoke the search functionality within the workflow to supplement it with up-to-date public information.
• Team space:Multiple people work together to maintain agents, workflows, knowledge bases, and plugins.
• Multi-channel distribution:It can be published on Tencent Yuanqi, the WeChat ecosystem, WeCom, and some partner platforms.
• API services:Create keys for published agents to embed in your own website, applications, or business systems.

Knowledge base and retrieval enhancement

Tencent Yuanqi’s knowledge base is used to store and manage external materials, and it retrieves the relevant sections so that the model can provide answers. The platform can automatically split documents, and it also allows for adjusting the way in which they are split based on segment identifiers or character length.

• It supports importing documents, Q&A pairs, web content, and structured data;
• Common formats include PDF, Word, PowerPoint, Excel, and images, etc.;
• It is possible to set tags, the expiration date of the knowledge, and the search range;
• Manual intervention can be applied to the parsing results and content segments;
• Multiple agent applications can share the same knowledge base;
• Semantic search can be used to find content that expresses different things but has a related meaning.

The successful upload of a file does not guarantee that its content can be parsed accurately. Scanned documents, complex tables, multi-page structures, and low-resolution images still require verification of their sections, numerical values, and heading levels.

Workflows and multi-agent systems

Visual workflow

Workflows break down complex tasks into adjustable nodes, making them suitable for tasks with well-defined processes and a need for consistent output. Developers can manage parameters, conditions, loops, batch data, plugins, as well as the relationships between code and large models.

Multi-Agent mode

Multi-agent applications are suitable for distributing tasks such as customer service, searching, writing, and data querying among different agents. The main agent selects the appropriate sub-agents based on the user’s intentions, thereby avoiding the need to include all knowledge and tools in a single prompt.

• Each sub-agent should have clear responsibilities and boundaries for refusing requests.
• Maintain consistent versions and permissions when sharing knowledge bases;
• Prevent conflicts that may arise when multiple agents handle the same intent at the same time;
• Test the forwarding rules, fallback in case of failure, and the final output;
• Keep a manual processing option available for high-risk actions.

Distribution via official accounts and WeChat ecosystem

The official account bot can be authorized to operate one or more official accounts, and it will automatically retrieve past articles to incorporate them into the knowledge management system. Once the prompt parameters and question-and-answer functions have been adjusted, the bot can be deployed in the official account services for which it has been authorized.

• Publish to the WeChat official account or service account backend;
• Add the agent mini-program to the official account’s menu or articles;
• Post to WeChat customer service for handling common inquiries;
• Publish on WeCom and other eligible partnership channels;
• Provide a web experience by sharing links or QR codes;
• Convert published agents into API services.

Unverified official accounts may require users to send a “Continue” message in order to receive responses that take some time to process. It is also not appropriate to design processes within such accounts that necessitate fans to upload documents in order to proceed.

Tencent Yuanqi prices and quotas

The official website of Tencent Yuanqi highlights free model support; the price of individual subscription plans is not disclosed. The new version of the Yuanqi API offers a free monthly quota, while enterprise-level capacity and support require use of the Tencent Cloud Intelligent Agent development platform, with pricing based on that platform’s dashboard.

1 million Tokens refer to the amount of model usage allowed, and they are not equivalent to a fixed number of conversations. Prompt text, context, knowledge snippets, workflows, and the length of the output all influence the actual amount of tokens consumed.

How to choose between the free version and the enterprise-grade solution

Tencent Yuanqi Usage Guide

Create the first conversational AI agent

1. Log in using your mobile phone number or a supported Tencent account and complete the real-name verification.
2. Enter the development platform and select to create a new conversational AI application.
3. Fill in the name, profile picture, biography, role prompt, and scope of capabilities.
4. Add a welcome message and example questions to guide users in asking proper questions.
5. Configure the knowledge base, plugins, workflows, or sub-agents as needed.
6. Test normal, blurred, unauthorized, and malicious issues in the preview area.
7. Select Public, Link Available, or a specified channel and complete the publication.

Create a WeChat official account bot

1. Select the official account AI agent and authorize one or more official accounts.
2. Check whether historical articles have been synchronized to the knowledge management module.
3. Set role responsibilities, specify prohibited response contents, and define rules for routing to a human agent.
4. Add proprietary knowledge, FAQs, and knowledge expiration dates.
5. Use the official account’s private messages to test the answers and the response time.
6. Select to publish a public account, and check the authentication details as well as any channel restrictions.

Publish the agent as an API

1. First, complete the configuration and debugging of the agent, and then release it officially.
2. Enable the API service in the distribution channel to obtain the AppID and AppKey.
3. Use Bearer authentication on the server side; do not include the AppKey in the web interface.
4. Send messages in accordance with the official interface format and handle streaming or complete responses.
5. Add timeout, retry, concurrency control, logging, and content security checks.
6. Monitor the monthly Token quota; switch to an enterprise-level platform once it exceeds the prototype scale.

Which users are it suitable for

• WeChat official account operator:Create content avatars and round-the-clock Q&A using historical articles.
• Corporate customer service team:Create robots for policies, products, after-sales service, and guidelines.
• Knowledge managers:Organize documents, Q&A, and databases into shared knowledge services.
• Operations and Marketing Staff:Use workflows to generate content, retrieve information, and process leads.
• Teachers and trainers:Create course Q&A, practice guidance, and resource search assistance.
• Product Manager:Quickly test the AI application prototype before starting formal development.
• Developer:Integrate your own products via plugins, MCP, and agent APIs.

Product advantages

• Zero-code configuration reduces the barriers to creating and debugging agents.
• Articles on official accounts can be turned into a knowledge base, and Tencent provides a complete distribution network.
• Knowledge, plugins, workflows, and multi-agent systems can be used in combination;
• Workflow nodes cover intent, logic, code, search, and batch processing;
• It supports extension methods such as MCP and OpenAPI;
• An API prototype quota of 1 million model tokens per month is provided;
• When production-grade capabilities are required, it can be integrated with Tencent Cloud Enterprise Platform.

Usage restrictions and precautions

• The service agreement is currently available only to users in the Chinese mainland, and real-name registration is required.
• The concurrent limit for the standard component API is 10, which makes it unsuitable for services with high traffic levels that have not been evaluated.
• The new version of Meta Components no longer offers the option to purchase additional API tokens; for more than that, an enterprise platform must be considered.
• Different distribution channels have their own separate authentication processes, review procedures, and limitations on interaction.
• The model may experience hallucinations, and knowledge retrieval might fail to retrieve relevant information or may cite outdated content.
• Complex workflows increase delays, failure points, and Token consumption;
• The platform may update, modify, or discontinue certain features; older applications need to be upgraded in accordance with the migration documentation.

Team permissions and data security

According to the official service agreement, members of a team space may have full editing rights over the agents, workflows, knowledge bases, and plugins available within that space. It is necessary to verify the trust level of potential members before inviting them, and care should be taken to avoid placing production keys and sensitive information in spaces with overly broad permission settings.

• Only members who are actually involved in the maintenance are invited to join the team space;
• Obtain legitimate authorization before uploading third-party personal information to the knowledge base;
• Prepare clear privacy notices for agents intended for end-users;
• The AppKey and channel authorization credentials are stored only on the secure server side;
• Create backups for deletions of knowledge, plugins, and workflows;
• For queries related to healthcare, law, finance, and government affairs, manual verification and referral to human agents are required.

After the component account is canceled, the developer data and user data will be deleted or anonymized in accordance with the relevant rules, and the intelligent agents that were created will no longer be able to be accessed. Services that are still in operation must notify their users first and arrange for a migration process.

API and open-source information

Tencent Yuanqi offers an official HTTP agent API that enables authentication using AppID and AppKey, allowing published applications to be integrated into other products. Regular APIs are suitable for prototypes and scenarios with moderate to low traffic levels; for enterprise-level solutions, it is advisable to use Tencent Cloud Agent development platform.

Tencent Yuanqi’s core platform, the Hunyuan model, and the related runtime services are not open-source projects; the other AI repositories available in Tencent’s GitHub account do not represent the source code for Yuanqi. Third-party SDKs or plugins require separate verification regarding their developers, maintenance status, and licensing terms.

Frequently Asked Questions

Is Tencent Yuanqi free?

The official website currently offers free services for creating agents and using models; the specific features and quotas are subject to those specified in the dashboard. Enterprise-level resources, cloud model calls, and high-concurrency services may incur costs.

What is the free quota for Tencent Yuanqi API?

The new version of MetaComponents offers 1 million model tokens per month, with no additional tokens available for purchase. The API allows for 10 concurrent requests; if more are needed, consideration should be given to Tencent Cloud Enterprise Platform.

Can public account articles be turned into a knowledge base?

Yes, once the official account is authorized, historical articles will be automatically retrieved and incorporated into the agent’s knowledge management system. Before publishing, it is still necessary to check the relevance of the articles, the results of their analysis, and any content that should not be made public.

Does Tencent Yuanqi support workflows and multiple agents?

It is supported; nodes such as conditions, loops, batch processing, plugins, code, and large models can be used, and multi-Agent applications can also be created. Complex processes require thorough debugging and error handling.

Is Tencent Yuanqi open-source?

No, the core platform and models belong to Tencent Online Services. The company provides APIs, plugins, and MCP extension capabilities, but it does not make the complete source code of its products available.