Skip to main content

Changelog

[2.4.6] - 2026/08/28​

  • New models: hy4-preview, deepseek/deepseek-v4-flash-vision-exp, glm-5.3-flash, deepseek-v4-flash-0731, deepseek-v4-pro-0813

[2.4.5] - 2026/08/14​

  • New models: glm-5.3
  • DeepSeek-V4 series model pricing adjusted with a peak/off-peak pricing mechanism, see announcement

[2.4.4] - 2026/08/03​

  • New models: mimo-v2.5-pro

[2.4.3] - 2026/07/17​

  • New models: kimi-k3

[2.4.2] - 2026/07/06​

  • New models: hy3, kimi-k2.7-code-highspeed
  • New image generation models: HY-Image-3.0-Plus-4090-Tob-v1.0, HY-Image-v3.0-I2I-ToB-v1.0.1
  • Reduced MiniMax-M3 model pricing, see announcement

[2.4.1] - 2026/06/24​

  • New models: glm-5.2, kimi-k2.7-code, hy-mt2-pro, hy-mt2-plus, hy-mt2-lite, hunyuan-role-latest, hy-role

[2.4.0] - 2026/06/18​

  • Resource-Point Plan now supports Token deduction
  • Token Resource Pack discontinued
  • CloudBase Agent and Builder now support using large models in the environment
  • Compatible with OpenAI Responses API

[2.3.2] - 2026/06/03​

  • New models: Qwen3.5-Flash, Qwen3.5-Plus, MiniMax-M3, DeepSeek-V4-Flash (direct from vendor), DeepSeek-V4-Pro (direct from vendor)
  • hy3-preview under the cloudbase provider now supports deducting Tokens gifted by the Mini Program Growth Plan

[2.3.1] - 2026/05/29​

  • Support Anthropic protocol
  • Mini Program Growth Plan model upgraded to hy3

[2.3.0] - 2026/05/08​

  • Support Token resource packages
  • Support models: deepseek-v4-flash, deepseek-v3.2, deepseek-v4-pro, glm-5, glm-5-turbo, glm-5.1, glm-5v-turbo, hy3-preview, kimi-k2.5, kimi-k2.6, minimax-m2.5, minimax-m2.7

[2.2.0] - 2026/03/08​

  • Support LLM call logs with request content, response content, latency, and first token response time
  • Support Agent analytics dashboard with active users, API latency, and success rate

[2.1.0] - 2026/02/04​

  • Support Google Aggent Development Kit
  • Supports Tencent Hunyuan and Coze Platform
  • Support online debugging

[2.0.0] - 2026/01/02​

  • Supports LangChain, LangGraph, and CrewAI development frameworks
  • Supports Tencent Hunyuan and Tencent Cloud Agent Development Platform
  • Support running Agent on both cloud function and cloud hosting platforms
  • Support both Javascript and Python

[1.14.0] - 2025/12/25​

  • Added Hunyuan text-to-image model

[1.13.0] - 2025/05/16​

✨ Core Features Upgrade​

Cross-Platform Voice Capability Enhancement for Agent UI

  • Support 4 scenario-based voice tone switching (chat/reading/consultation/general)
  • Added support for playback in mute mode
  • Mini Program & low-code components support voice playback speed adjustment
  • Optimize the display logic for dual titles (hide redundant titles in the content area)
  • Resolve jitter issues in mixed image-text layouts

Agent UI Web Component Capability Expansion

  • Added voice playback feature
  • Support three types of tool cards such as map/weather
  • Fixed several interaction issues

🌐 Extended Open Capabilities​

MCP Server Template Expansion

  • Tencent Hunyuan 3D MCP: Supports generating 3D models from text/images (MAX/FBX/GLB formats)
  • MarkItDown Service MCP: Enables any file → Markdown format conversion
  • CRIC Property AI MCP: Integrates dedicated intelligent assistant capabilities for the property management industry

Functional Agent Expansion

  • LKE Functional Agent Template: Supports integration with large model knowledge engines

🚧 Upcoming Features Preview

  • The Agent interface supports users in autonomously switching models
  • Optimized the effectiveness of knowledge base search and enhanced compatibility with multiple file formats
  • WeChat official accounts/service accounts/WeChat Customer Service integration with Functional Agent capabilities
  • Agent supports integration with more third-party Agent development platforms
  • Supports one-click conversion of OpenAPI to MCP
  • Provides MCP for integrating with third-party knowledge base search and Agent platforms

The feature will be rolled out in batches. For the specific schedule, please refer to the version update notifications.

[1.12.0] - 2025/04/29​

✨ Core Features Upgrade​

Agent supports voice input/output capabilities

Add native voice input/output capabilities to Agent, enabling developers to quickly build multimodal interaction scenarios:

Agent supports multi-session capability

Supports a single Agent instance concurrently handling multiple independent sessions, with each session maintaining separate history

Agent UI Mini Program components support tool cards

Agent UI mini program source code supports ToolCard tool cards, supporting flexible customization of MCP display effects. Supports custom business logic such as product recommendation or map navigation

⚙️ Developer Experience Improvements​

File Upload Component Security Upgrade

Due to the security policy upgrade of the WeChat Mini Program base library, the file upload feature parsing process may be affected. Developers are advised to upgrade the Agent UI components.

  • Mini-program source code components, please update and replace to the latest version
  • For WeDa low-code components, please create a new page and then drag to update the agent-ui block again. After completion, publish the application.

🚧 Upcoming Features Preview

  • The Agent interface supports users in autonomously switching models
  • Optimized the effectiveness of knowledge base search and enhanced compatibility with multiple file formats
  • WeChat official accounts/service accounts/WeChat Customer Service integration with Functional Agent capabilities
  • Agent supports integration with more third-party Agent development platforms
  • Supports one-click conversion of OpenAPI to MCP
  • Provides MCP for integrating with third-party knowledge base search and Agent platforms

The feature will be rolled out in batches. For the specific schedule, please refer to the version update notifications.

[1.11.0] - 2025/04/17​

✨ Core Features Upgrade​

  • Supports barrier-free cloud-based hosting for MCP Server

    Fully supports MCP Server written in mainstream languages such as Java/Python/GoLang/PHP, without invasive framework modifications.

    Local MCP Servers can be seamlessly migrated to TCB hosting, automatically gaining access to secure and reliable remote MCP services.

🌐 Extended Open Capabilities​

  • TCB MCP market officially launched

    TCB provides a one-stop MCP Server development and hosting platform, supporting rapid publishing and deployment of MCP Server.

    Added 30+ standardized service modules, covering mainstream development scenarios:

    DomainRepresentative Service Modules
    AI EnhancedTCB AI Capabilities / Memory Module / Meta Plugin / Puppeteer Automation
    Data ServicesMongoDB/MySQL/Postgres/Redis Database Access / COS File & Data Processing
    Developer ToolsGithub/Gitee/Gitlab Code Management / Figma Design Collaboration / Apifox API Debugging
    Maps & Lifestyle ServicesTencent Location Service / Amap / Caiyun Weather / ALAPI URL Shortener & Data Hotlist
    Productivity ToolsNotion Knowledge Base / QuickChart Visualization / Mermaid Chart Rendering / Zapier Universal Workflows
    Cloud-Native InfrastructureTCB Environment Management / Lighthouse Cloud Server / CNB Cloud-Native Build / Trends Hub Hotspot Analysis

🚧 Upcoming Features Preview

  • Agent supports session feature
  • Agent voice input and output features are now available
  • The Agent interface supports users in autonomously switching models
  • Optimized the effectiveness of knowledge base search and enhanced compatibility with multiple file formats
  • WeChat official accounts/service accounts/WeChat Customer Service integration with Functional Agent capabilities
  • Agent supports integration with more third-party Agent development platforms
  • Supports one-click conversion of OpenAPI to MCP
  • Provides MCP for integrating with third-party knowledge base search and Agent platforms

The feature will be rolled out in batches. For the specific schedule, please refer to the version update notifications.

[1.10.0] - 2025/03/28​

✨ Core Features Upgrade: Full Support for MCP​

  • Agent supports the MCP protocol
    Agent can integrate MCP Servers, enabling AI to automatically select tools based on task requirements and complete complex operations through multiple tool invocations.

  • TCB supports the development and hosting of MCP services
    Provides a one-stop service from development to deployment, enabling you to host and build your own MCP services on TCB. Provides 6 built-in MCP Server templates

    • TCB AI Capabilities: Provides TCB AI atomic capabilities through the MCP protocol, such as knowledge base retrieval, database querying, large model invocation, web search, image parsing, etc.
    • Yuanqi Plugins: Provides built-in access to 19 official Tencent Yuanqi plugins, supporting capabilities such as image generation, audio generation, and Sogou search. - Yuanqi Agent: Converts Tencent Yuanqi Agent into MCP plugins for use, enabling integration with personal WeChat official account articles (Yuanqi Agent features a built-in capacity of 100 million tokens).
    • Memory Module: Based on TCB database + knowledge graph, it can provide memory for large models and supports storing user-level information.
    • Puppeteer: Can control the Chrome browser to perform operations
    • Blank Template: Developers can quickly implement their own MCP Servers, enabling Agents to connect to their own data and features.

  • Agent UI Component Upgrade

    • Mini Program Agent supports invoking shared TCB environments
    • Added status display for the MCP tool invocation process

🚧 Upcoming Features Preview​

  • Quickly import open-source MCP Servers
  • Agent supports session feature
  • Agent voice input and output features are now available
  • The Agent interface supports users in autonomously switching models
  • Optimized the effectiveness of knowledge base search and enhanced compatibility with multiple file formats
  • WeChat official accounts/service accounts/WeChat Customer Service integration with Functional Agent capabilities

Please stay tuned for further announcements regarding the specific launch date.

[1.9.0] - 2025/03/21​

✨ Core Features Upgrade​

  • Agent supports image upload, parsing, and Q&A
    Added support for uploading images during conversations to enable parsing and AI-generated responses, enhancing multimedia interaction capabilities.
  • Agent supports multiple knowledge base selection
    Supports selecting 3 knowledge bases during conversations for comprehensive search, enhancing answer accuracy and information coverage.

⚙️ System Improvements​

  • Enhanced Knowledge Base Upload and Management Features
    Supports drag-and-drop file upload for streamlined operations Supports batch upload of up to 200 files Added a batch file deletion feature to improve management efficiency.

  • Error Handling and Notification Optimization
    Enhanced error message prompts provide clear guidance for situations such as Token expiration and service overlimit. Detects the WeCom environment and displays compatibility prompts

  • Enhanced Dialogue Processing Logic
    Optimized the dialogue logic for the WeChat Official Account Platform and fixed the exception of potentially replying to the same question repeatedly. Refactored the file content parsing interface to resolve timeout issues in complex file parsing.

🌐 Extended Open Capabilities​

🚧 Upcoming Features Preview​

  • Agent autonomously invokes external tools to implement complex business logic
  • Agent supports session feature
  • Agent voice input and output features are now available
  • The Agent interface supports users in autonomously switching models
  • Optimized the effectiveness of knowledge base search and enhanced compatibility with multiple file formats
  • WeChat official accounts/service accounts/WeChat Customer Service integration with Functional Agent capabilities

The feature will be rolled out in batches. For the specific schedule, please refer to the version update notifications.

[1.8.0] - 2025/03/14​

✨ Core Features Upgrade​

  • Agent supports integration with TCB data models
    Added support for integration with TCB data models to enable querying structured data through natural language and generating AI responses.

  • Agent UI feature optimization, officially open-sourced

⚙️ System Improvements​

  • Integrated TCB extension capabilities for WeChat DevTools IDE Developers can view AI+ usage statistics and develop Agent in WeChat IDE via TCB - extension features.

  • Large Model Usage Statistics Feature Launched
    You can view specific token usage via Development Platform - Package Usage - Resource Dashboard.

  • Agent file upload supports management switch
    Due to the additional large model token consumption caused by file uploads during conversations, a file upload management switch has been added to the Agent. This feature is only available for Entry Plan and above. After the large model resource package is launched, this capability will be made available to the Individual Plan.

  • Optimized Dialogue Processing Logic

    • WeChat automatically trims leading and trailing whitespace characters in conversations
    • The query scope for cloud-based conversation history has been optimized to the last 1 day, enhancing the relevance of dialogue context.

🌐 Extended Open Capabilities​

🚧 Upcoming Features Preview​

  • Agent autonomously invokes external tools to implement complex business logic
  • Agent supports image upload in the dialogue interface.
  • Agent supports session feature
  • Agent voice input and output features are now available
  • The Agent interface supports users in autonomously switching models
  • Agent has supported integration with multiple knowledge bases
  • Optimized the effectiveness of knowledge base search and enhanced compatibility with multiple file formats
  • WeChat official accounts/service accounts/WeChat Customer Service integration with Functional Agent capabilities

The feature will be rolled out in batches. For the specific schedule, please refer to the version update notifications.

[1.7.0] - 2025/03/07​

✨ Core Features Upgrade​

  • Agent UI document context supports file upload
    Added support for users to upload documents as conversation context during dialogues (currently available on: Mini Program source code components and visually built Web applications; adaptation for visually built Mini Programs is under development), expanding multimodal interaction scenarios.

  • TCB Environment Resource Sharing
    Supports the same entity sharing the TCB environment to invoke AI and related resources

  • Optimized the user experience of Mini Program Agent UI components Supports selecting and copying text, the input box dynamically adjusting its height based on text content, and exposes more component property configurations.

🌐 Extended Open Capabilities​

  • Functional Agents provide standard interface documentation The Functional Agent framework provides the IBot interface. Developers only need to implement a class that satisfies IBot to deploy the corresponding Agent. For details, refer to the documentation

  • Agent API Open Invocation
    Developers can invoke Agent interfaces via API Keys to achieve automated integration and third-party service connectivity.

  • Custom UI Output Solution
    Added the capability for Agents to return custom UI structures. Developers can quickly implement personalized interaction designs via the Guide Documentation.

⚙️ System Improvements​

  • Optimized Agent Dialogue Effectiveness
    Improved the accuracy of web search results, optimized the weight allocation policy for setting information, and enhanced context understanding and response relevance.

  • Error Notification and Process Optimization
    Optimized the experience during the Agent building process.

🚧 Upcoming Features Preview​

  • Integrate with TCB database as a structured data source, enabling direct citation in conversations.
  • Agent supports image upload in the dialogue interface.
  • Agent supports session feature
  • Agent voice input and output features are now available
  • The Agent interface supports users in autonomously switching models
  • Agent has supported integration with multiple knowledge bases
  • Optimized the effectiveness of knowledge base search and enhanced compatibility with multiple file formats
  • WeChat official accounts/service accounts/WeChat Customer Service integration with Functional Agent capabilities
  • Supports viewing large model invocation usage statistics
  • Provide out-of-the-box Functional Agent templates and solutions

The feature will be rolled out in batches. For the specific schedule, please refer to the version update notifications.

[1.6.0] - 2025/02/27​

✨ Core Features Upgrade​

  • Release of Functional Agent
    Added support for Functional Agents that enable complex logic development, compatible with open-source Agent frameworks, and capable of integrating with external APIs and diverse data sources to enhance multi-scenario scalability.
  • Hunyuan Turbo S Fast Thinking Model Launch
    Added the "Hunyuan Turbo S" fast-thinking model, delivering responses within seconds (token output speed increased by 100%, first-token latency reduced by 44%), complementing existing slow-thinking models like DeepSeek-R1.

🎨 Interactive Experience Optimization​

  • Agent UI Conversation Interface Refresh
    • Refactored Agent UI, optimized animation feedback for the thinking state and online interactions
    • Conversation History: Agent UI has added a conversation history display feature that supports backtracking queries.
    • Online Toggle: C-end users can independently disable the online search feature.
    • Bug Fixes: Resolved the rendering issue with bold text and line breaks in Markdown.

🌐 Extended Open Capabilities​

  • Knowledge Base Management API
    Developers can use APIs to achieve automated management of knowledge base documents, supporting features such as creation, deletion, and batch operations. Refer to the documentation.

⚙️ System Improvements​

  • Optimized Answering Logic
    Fixed the exceptions of repeating user questions and misusing Markdown formatting in Agent responses, enhancing output accuracy.

🚧 Upcoming Features Preview​

  • Supports uploading documents and images in the dialogue interface for conversations.
  • Supports voice input and output features
  • The Agent interface supports users in switching models
  • Integrate with TCB database as a data source, enabling direct reference of structured data in conversations.
  • knowledge base supports automatic synchronization of WeChat official accounts
  • Optimized the effectiveness of knowledge base search
  • knowledge base supports additional file formats
  • Supports invoking Agent APIs via API key
  • Supports integration of WeChat official accounts / service accounts / WeChat customer service with functional Agent

The feature will be rolled out in batches. For the specific schedule, please refer to the version update notifications.

[1.5.0] - 2025/02/21​

✨ Core Features Upgrade​

  • Web Search Capability
    Agent now supports large models like DeepSeek in accessing real-time internet data search.
  • Dynamic Model Management
    Added model version control feature, supports specifying different versions of large models such as DeepSeek-V3

⚡ Performance Improvements​

  • Document Processing Capability Enhancement
    The file size of knowledge base has been expanded to 100MB (Markdown files are limited to 10MB).

🌐 Open Interface​

  • Full-scenario API Support
    Provides HTTP REST API and OpenAI SDK-compatible interface invocation capabilities

[1.4.0] - 2025/02/21​

🧠 Model Ecosystem​

  • New Model Access
    Officially supports large language models including DeepSeek R1 and DeepSeek-V3

🚀 Application Building​

  • Quick Deployment Tool
    Provides an end-to-end feature for AI application deployment from conversation flow construction to deployment.
  • Interactive Experience Enhancement
    The console has added a real-time preview module for large language model dialogues.

📱 Cross-Platform Support​

  • Mini Program Integration Solution
    Launches the Agent UI component library and industry template suite.

[Historical Versions]​

1.3.0​

  • Architecture Upgrade
    WeChat Mini Program base library integrates native AI SDK

1.2.0​

1.1.0​

  • Extended Development Capabilities
    • SDK adds toolchain invocation support
    • Seamless integration of AI capabilities into cloud functions

1.0.0 - Initial Release​

  • Foundational Capability Building
    • Hunyuan large model API integration
    • Visual Agent Construction Framework Multi-platform publishing capability within the WeChat ecosystem (Mini Program/WeChat official accounts/WeCom/customer service systems)
    • RAG Knowledge Base Engine (Supports Document Q&A Scenarios)