3Chat.ai v5.1 Release Notes: Enhanced Agent Builder Experience, Agent Version Management, Enterprise WeChat Group Chat Support, and Upgraded Inbox View

Zhiyu · June 7, 2026, 6:51am

Hi everyone,

3Chat.ai v5.1 is officially released.

In v5.0, we officially launched the Agent Builder, helping merchants build their own 3Chat Agents using natural language.

In v5.1, we have further upgraded around Agent building experience, Agent controllability, group chat reception capabilities, channel capabilities, and inbox efficiency.

Key highlights of this update include:

Enhanced interaction and guided experience for Agent Builder
Version management support for Agent Builder / Guidance / Skills
Main Agent now supports WeCom group chat scenarios
Main Agent now supports web search capabilities
Personal WeChat now supports reading voice messages and triggering manual handover from mobile
Inbox adds Work View, Channel View, and Custom View
Optimized display of knowledge base processing progress
Added space information to manual handover emails

Below is the detailed update guide for this version

1. Agent Builder Experience Upgrade

1. Clearer Guided Setup, Smoother Generation Process

In v5.0, the Agent Builder already allowed users to complete the initial setup of an Agent using natural language.

In v5.1, we further optimized the Builder’s interaction experience and guidance flow, making it easier for users to understand:

Which stage of the setup the current Agent is in
What business information needs to be added next
How to test the Agent after generation is complete
Where to go to connect channels after successful publication

This optimization includes:

Optimized Builder status tags for more accurate display of setup status
Optimized guidance for the first question and business goal questions
Optimized progress display to help users clearly see which steps have been completed
Automatic display of the sandbox testing area upon successful generation for immediate Agent testing
Added channel connection guidance action cards after the first successful publication
Optimized interaction details such as action cards, input fields, and Markdown display

This means that when you use the Builder to create an Agent, the entire process feels more like a clear setup wizard rather than just chatting with AI.

2. Agent Builder Supports Version Management

In v5.1, Agent Builder adds version management capabilities.

Starting from this version, every time you complete an overall publication via the Builder, a formal version is created. You can now clearly see which Builder publication generated the currently live Agent.

Version management supports:

Draft versions
Current live version
Historical live versions
Viewing historical versions
Creating or overwriting the latest draft from a historical version

Important note:

When you choose a historical version to restore, the system will not directly roll back the live data. Instead, it will create or overwrite the current latest draft based on the selected historical version. You can review, modify, and test it first before publishing it to the live environment.

This approach helps avoid accidental operations directly affecting live customer reception and gives the team a chance to confirm before publishing.

3. Guidance and Skills Also Support Independent Version Management

In addition to the Agent Builder itself, two core configurations of the 3Chat Agent have added version management:

Guidance
Skills

This means you can view the historical versions of guidance and skills separately, and restore to a draft based on a historical version before deciding whether to publish.

Suitable for the following scenarios:

Comparing historical versions after modifying the Agent’s response style
Finding that the effect of a modified Skill is not as good as the old version
Knowing who published what modification and when during team collaboration
Wanting to keep a stable version while continuing to optimize the next version of the Agent in the draft

4. Boundaries of Using Builder Versions

To avoid misunderstandings, it is important to clarify a key boundary:

The version records of the Agent Builder mainly track content generated and published with the Builder’s participation.

If you manually create a Skill outside of the Builder, this Skill will not immediately be included in the Builder’s version records. The Builder will only consider including this content in the new generation results based on the current business context, as well as existing Guidance and Skills, the next time you regenerate or modify the Agent via the Builder.

Therefore, if you want the setup, modification, and restoration of your Agent to be more unified, we recommend completing the continuous setup and adjustment of the Agent primarily through the Builder.

If you have overall restoration needs later, it is also recommended to rely on the Builder’s versions.

2. Main Agent Capability Expansion

1. Supports WeCom Group Chat Scenarios

In v5.1, the Main Agent begins to support standard group chat scenarios.

This group chat capability initially supports WeCom scenarios. Personal WeChat group chats are not yet supported and are expected to be launched in v5.4.

The situations an Agent faces in group chats are much more complex than in one-on-one chats. A group may simultaneously contain customers, colleagues, multiple accounts belonging to your organization, and multiple parallel consultations.

v5.1 provides systematic support for these scenarios.

2. Agent Can Distinguish Between Contacts, Colleagues, and Own Channel Accounts

In WeCom group chats, the Main Agent can now distinguish between three types of identities:

Contact: i.e., the customer
Colleague: Internal enterprise members
Own Channel Account: Accounts connected to 3Chat and managed by the Agent

For customer messages, the Agent determines which account the customer wishes to be served by based on whether the customer @s the own channel account or quotes/replies to a message from the own channel account.

For colleague messages, the Agent references these messages as group chat context but will not reply to colleagues in any scenario.

In other words, colleagues can supplement background information in the group without worrying about the Agent mistakenly replying to them as if they were customers.

3. One Own Account Can Simultaneously Handle Multiple Customers

In group chats, one own channel account can simultaneously handle consultations from multiple customers.

For example, in the same WeCom group:

Customer A is asking about product prices
Customer B is asking about shipping times
Customer C is asking about after-sales policies

As long as they each trigger the Agent, the Main Agent can assemble context based on different contacts and handle the corresponding issues separately.

When one contact triggers a manual handover, the Agent can continue to handle messages from other contacts. It will not stop serving the entire group just because one customer has been handed over to manual support.

4. Supports Manual Handover Processing in Group Chats

Manual handover has also been adapted for group chat scenarios.

When a specific contact triggers a manual handover, only that contact enters manual processing, while other contacts can continue to be replied to by the Agent.

Additionally, if the managed own account actively replies to messages from the mobile end, the system recognizes that the channel account has been taken over by a human. At this point, the account enters a more complete manual takeover state, preventing AI and humans from interrupting simultaneously under the same account identity.

This makes AI reception in group chats closer to real-world business collaboration: AI can handle most consultations, while humans only intervene for specific customers or risk scenarios.

5. Deployment, Contact Groups, and Skills Support Group Chat Conditions

To allow for more granular configuration of group chat scenarios, v5.1 also extends related configuration capabilities.

Now, in deployment and contact group configurations, you can use more group chat-related information, such as:

Group information
Group member information
Session information
Group chat context

When identifying customers, the Main Agent no longer only reads individual contact profiles but can read more complete contact context.

Simultaneously, Skills and condition adherence have added read/write capabilities for groups and group custom fields. You can design more refined automation rules around group chat business scenarios.

6. Main Agent Supports Web Search

In v5.1, the Main Agent adds web search capabilities.

When your business requires the Agent to query real-time information or assist in answering based on public web page information, you can use the web search tool.

How to use:

Enable “Web Search” in the basic configuration of the 3Chat Agent
In the tool call location within the Skill, select “Web Search”
When the Skill is triggered, the Agent can call web search to assist in completing the answer

Please note:

Web search is currently only supported on the 3Chat Commercial Edition
It is disabled by default and must be manually enabled in the basic configuration
After enabling, you must specify the tool call in specific Skills for the Agent to use web search for corresponding tasks

This allows the Agent to not only rely on content in the knowledge base but also obtain more real-time information in specific business scenarios.

3. Channel Capability Expansion

1. Personal WeChat Supports Reading Voice Messages

In v5.1, the Personal WeChat channel supports reading voice messages.

When customers send voice messages via WeChat, 3Chat can read the voice content and allow the Agent to continue understanding the customer’s question based on the voice message.

This is very important for many WeChat private domain scenarios. Customers do not always type out their inquiries; voice messages are very common, especially in pre-sales communication, after-sales feedback, and repeat purchases by old customers.

With voice message support, the reception experience for the Personal WeChat channel becomes more complete.

2. Personal WeChat Supports Triggering Manual Handover via Mobile Messages

In v5.1, Personal WeChat also supports triggering manual handover by sending messages from the mobile device.

When a human agent directly replies to a customer from the mobile end, the system recognizes that the current session has been intervened by a human and switches to manual processing mode.

This prevents AI from continuing to auto-reply after a human has taken over, reducing the disturbance caused by AI and humans replying to the customer simultaneously.

Suitable for the following scenarios:

Sales personnel temporarily taking over customers via mobile
Customer questions requiring human judgment
Important customers needing direct human communication
Humans continuing to follow up and close deals after initial AI reception

4. Other Optimizations and Bug Fixes

1. Added Space Name to Manual Handover Emails

To facilitate multi-space users in managing manual handover notifications, v5.1 adds space information to manual handover emails.

When you receive a manual handover email, you can directly see which space the notification comes from.

For users managing multiple brands, stores, or business teams simultaneously, this helps you quickly determine which space to enter to handle customer issues.

2. Added Banner Prompt for Knowledge Base Processing Progress

When there are files or web pages being processed in the knowledge base, a processing progress banner is displayed at the top of the page.

You can directly see the overall progress of the current knowledge base processing queue, for example:

Processing Progress: 3/10

This means:

3 represents the number of files or web pages that have been successfully processed
10 represents the total number of files or web pages currently in the queue to be processed

If you continue to upload files or crawl web pages while existing files are still being processed, the new content will be appended to the current processing queue, and the total number will increase accordingly.

Once all content processing is complete, the banner will automatically disappear.

This allows you to clearly see whether the knowledge base has finished updating and determine if the Agent can already use the latest knowledge content.

3. Inbox Adds Three New Views

v5.1 upgrades the inbox views to help you filter and handle sessions more efficiently.

Three new views are added:

Work View

The Work View is primarily for users who need to reply to messages in the inbox.

It categorizes sessions based on their current processing status, including:

All Messages
AI Processing
Pending Manual Handover
Assigned to Me
Closed

You can quickly find sessions that need manual processing and see which sessions are still being handled by AI.

Specifically:

“AI Processing” is used to view sessions still being handled by the Agent
“Pending Manual Handover” is used to view sessions that have triggered manual handover but have not yet been taken over by a member
“Assigned to Me” is used to view sessions assigned to the currently logged-in member
“Closed” is used to view sessions that have ended

Channel View

The Channel View categorizes sessions by connected channel.

For example:

LiveChat
Email
Instagram
Facebook
Personal WeChat
Sandbox Testing

If you have connected multiple channel accounts, you can also quickly enter sessions under a specific channel account from the Channel View.

This is very helpful for multi-channel operations. You can view customer inquiries by channel source and quickly locate customer messages under a specific account.

Custom View

The Custom View is used to meet the team’s own filtering habits.

You can create your own views based on business needs, such as:

High-intent customers
Need quotation
Not replied to in the last 24 hours
Customers from a specific channel source
Sessions handled by a specific customer service representative
Group chat / One-on-one chat type

This way, each team can organize their inbox according to their own working methods, rather than being limited to fixed lists.

5. Summary

3Chat.ai v5.1 is a version upgrade centered around “better Agent building, better management, better group chat reception, and better human collaboration.”