Hello everyone, 3Chat.ai v5.2 is officially released.
In v5.0, we launched the Agent Builder, allowing merchants to quickly build their own 3Chat Agents using natural language.
In v5.1, we further enhanced the version management capabilities for Agent Builder, Guidance, and Skills, making the building, modification, publishing, and rollback of Agents more controllable.
In v5.2, we continue to complete the key closed-loop for Agent Builder:
From Agent configuration generation to simulation testing, performance evaluation, continuous optimization, and version management, the Agent Builder is evolving from a “building tool” into an “Agent operations assistant.”
At the same time, this version also brings upgrades to the 3Chat Agent’s knowledge base, group chat configuration, user insights, and WhatsApp Business Account integration.
Below are the detailed update notes for this version 
I. Core Release: Agent Builder Supports Simulation Testing
1. From “Generating Agents” to “Automatically Verifying Agents”
In v5.2, Agent Builder officially supports simulation testing.
In the past, after users generated an Agent via the Builder, they often had to enter the sandbox themselves, manually扮演 as customers, and verify the Agent’s performance round by round.
Now, the Builder can proactively generate simulation test cases during the building process and guide you to run tests directly through natural language conversation.
This means you are no longer just “generating an Agent,” but you can also let the Builder help you verify:
- Whether the Agent can handle core consultation scenarios
- Whether the Agent can correctly guide customers
- Whether the Agent can collect customer information as expected
- Whether the Agent can transfer to human agents in appropriate scenarios
- Whether the Agent can achieve key business goals
- Whether the modified Agent is more stable than before
This is also the most important upgrade in v5.2: The Agent Builder begins to possess closed-loop capabilities from building to verification.
2. After the First Agent Generation, the Builder Proactively Recommends Test Cases
When you first generate a 3Chat Agent via the Builder, the Builder will proactively generate a batch of simulation test cases.
These cases will be displayed in the “Simulation Test” window on the right.
The system will prioritize recommending core process cases most suitable for the current Agent, such as:
- High-intent customer consultations
- Customers asking about prices or solutions
- Customers hesitating or观望
- Customers requesting after-sales service or guarantees
- Customers needing to be guided to leave contact information
- Customers needing human agent assistance
The Builder will recommend the most worthwhile test scenarios to run first based on the current Agent’s business positioning, Guidance, Skills, and core goals.
3. Click Run to Automatically Execute Multi-Round Simulation Tests
Users can directly click on the cases recommended by the Builder to run simulation tests. They can also converse directly with the Builder, asking it to think and find suitable cases to start running.
After running, the system will automatically simulate multi-round conversations between the simulated customer and the Main Agent.
During the process, you can see:
- Messages from the simulated customer
- Replies from the Main Agent
- The Agent’s processing process
- Agent Logs
- Current test running status
After the run is complete, the system will display the test results to help you judge whether the Agent met the expectations for that scenario.
It is recommended that users review the simulation test results and the final evaluation provided by the Builder. If there are deviations from actual scenarios, send the modifications directly to the Builder, and let the Builder help you make the changes.
If the test fails, the system will explain the differences between the Agent’s actual performance and the passing criteria, and provide optimization suggestions.
4. Recommended Cases Become Your Own Test Cases After Running
When you first run a case recommended by the Builder, this case is considered “adopted” by you.
Adopted cases will enter your simulation test case list, allowing you to continue viewing their detailed information, including:
- What scenario this case tests
- The background of the simulated customer
- Which Agent capabilities this case aims to verify
- Whether the current run result passed or failed
- Historical run records and conversation details
This ensures that test cases are not just one-time testing tasks but can be accumulated as a foundation for subsequent regression testing.
When you modify the Agent, you can continue to run these cases to verify whether the Agent can still stably handle core scenarios.
5. The Builder Continuously Recommends New Test Cases
In the recommended cases area, the Builder will continuously recommend new test scenarios for you.
Cases already adopted by you will not repeatedly appear in the recommendation list. The Builder will combine the current Agent’s capabilities and existing test records to continue recommending core scenarios that are more worthwhile to test.
This makes simulation testing feel more like a continuous optimization process:
- First, test the most important core paths
- Then, supplement with more typical customer scenarios
- Then, add custom scenarios based on actual business needs
- Finally, form a test case library specific to your business
6. Supports Custom Maintenance and Creation of Test Cases
For already adopted test cases, users can maintain them according to their actual business needs.
You can:
- Modify the test scenario name
- Modify the customer opening statement
- Supplement customer profiles
- Adjust customer background information
- Set initial values for contact fields
- Modify evaluation criteria
- Copy existing cases
- Create new custom cases
- Delete cases that are no longer needed
This means you can not only use test scenarios automatically generated by the Builder but also organize real customer consultations you have encountered into test cases.
For example:
- “The customer only asks for the lowest price and is unwilling to provide requirements”
- “The customer keeps asking if the results can be guaranteed”
- “The customer is very interested but needs to communicate with partners first”
- “The customer asked about after-sales issues, but it is actually a high-intent lead”
- “The customer needs human agent assistance, but the AI cannot make direct promises”
These can be organized into simulation test scenarios to continuously verify the Agent’s real reception capabilities.
7. What is the Difference Between Simulation Testing and Sandbox Testing?
In v5.2, the Builder provides both “Simulation Testing” and “Sandbox Testing” methods.
They are suitable for different scenarios.
Simulation Testing is more suitable for verifying core and typical scenarios.
These test scenarios are usually paths that must be verified before the first launch and serve as the foundation for subsequent regression testing. It is recommended to let the Builder automatically run simulation tests to systematically check the Agent’s performance.
Sandbox Testing is more suitable for users to扮演 as contacts themselves and manually experience conversation details.
If you want to test edge cases, temporary issues, or consultations that are not easily covered by standard scenarios, you can use sandbox testing.
In short:
Simulation testing is used for systematic verification, while sandbox testing is used for manual experience of details.
II. Agent Builder Current Capability List
With the release of v5.2, the Agent Builder is no longer just an “entry point for generating Agents” but covers the complete process from building, refining, testing, optimizing, to publishing.
Currently, the Builder supports the following capabilities:
1. Business Information Collection
Upon first entering the Builder, the system will collect key business information through natural language conversation, including:
- Industry
- Agent role positioning
- Product or service scope
- Core business goals
If information is incomplete, the Builder will continue to ask; once sufficient, it will proceed to solution generation.
2. Deep Guidance
The Builder will determine if additional key business details need to be supplemented based on the collected information.
For example:
- Whether to confirm target customer types
- Whether to confirm conversion goals
- Whether to confirm how to handle sensitive questions
- Whether to confirm which scenarios must transfer to human agents
This ensures the generated results are closer to real business needs, rather than just generating a generic Agent based on a few simple descriptions.
3. Agent Solution Generation
The Builder can automatically generate a complete Agent solution based on the business background, including:
- Role settings
- Communication style
- Reply strategies
- Condition following
- Human transfer rules
- Customer profile fields
- AI Skills
- Initial knowledge content
After generation, users can directly test, modify, or publish.
4. Solution Modification and Incremental Optimization
Users can tell the Builder what they want to modify using natural language.
For example:
- “Make the opening statement warmer”
- “Add a skill for handling price objections”
- “Do not directly promise results; instead, guide users to consult advisors”
- “Add this after-sales process to the Agent”
- “Make it more proactive in guiding customers to leave phone numbers”
The Builder will make incremental modifications based on the existing solution, rather than regenerating the entire Agent each time.
5. Automatic Knowledge Base Generation
The Builder can automatically organize FAQ knowledge bases based on the current business solution and upload them to 3Chat.
Suitable for generating:
- Frequently Asked Questions (FAQs)
- Product introductions
- Service descriptions
- Pricing information
- After-sales rules
- Franchise policies
- Purchase processes
- Appointment processes
This content can be retrieved and used by the Agent in subsequent replies.
6. Uploading Materials to the Knowledge Base
If you already have existing materials, you can upload them to the Builder for processing.
For example:
- Product manuals
- Service descriptions
- SOP documents
- Investment policies
- After-sales rules
- FAQ documents
- Event descriptions
The Builder can save these materials to the knowledge base for the Agent to use in subsequent conversations.
Note the following:
- If you say “save this material to the knowledge base,” the Builder will store it
- If you say “optimize the solution based on this material,” the Builder will attempt to understand and modify the current Agent
- If you upload materials without specifying their purpose, the Builder will first ask how you want to handle them
7. Optimizing Agents Based on Real Conversations
Once the Agent is running online, you can also provide real conversation records or screenshots to the Builder for analysis.
The Builder can help you determine:
- Where the Agent’s replies were insufficient
- Which questions require knowledge base supplements
- Which skills need adjustment
- Which scenarios should have new human transfer rules added
- Which scripts affect conversion
- Which customer intents were not correctly identified
The Builder will first provide optimization suggestions. After your confirmation, it will execute the modifications.
8. Simulation Testing
The new simulation testing capability in v5.2 allows the Builder to generate and run core scenario test cases, automatically evaluating Agent performance.
It is suitable for:
- Pre-launch verification
- Regression testing after modifications
- Core process acceptance
- High-intent customer path testing
- Human transfer scenario testing
- Information collection scenario testing
This is the most important new capability for the Agent Builder in this version.
9. Version Management
The Builder, Guidance, and Skills now support version management.
Users can distinguish between draft versions and official versions, view historical versions, and restore drafts based on historical versions before deciding whether to publish online.
This makes the continuous optimization of Agents safer, traceable, and rollback-capable.
10. Publishing to Online Channels
Once the Agent passes testing and the results are confirmed, it can be published to the online environment and integrated into actual channels to serve real customers.
The Builder will continue to help you continuously optimize the Agent based on real operational results, test results, and user feedback.
III. New Features for 3Chat Agents
1. Knowledge Base Adds “Smart Slicing”
In v5.2, the knowledge base slicing strategy adds Smart Slicing.
Compared to fixed segmentation methods, Smart Slicing processes document hierarchy and content boundaries more reasonably based on file format and content structure.
It strives to keep more relevant content within the same slice, thereby improving the completeness of knowledge retrieval.
We recommend prioritizing Smart Slicing, especially for the following materials:
- Product manuals
- Service description documents
- FAQ documents
- After-sales rules
- Operational SOPs
- Investment policies
- Long documents with multi-level structures
For Agents, the quality of knowledge base slicing directly affects whether complete context can be retrieved during answers.
The goal of Smart Slicing is to reduce situations where “only half the content is retrieved” or “context is split,” making it easier for the Agent to answer based on complete information.
2. Group Chat Scenario Configuration Continues to Expand
v5.2 continues to enhance configuration capabilities in group chat scenarios.
New and enriched configurations for group chat entity fields have been added in deployment and contact group settings.
This means you can configure more precise Agent scope and processing rules around group chat scenarios.
Additionally, in “AI Skills” and “Condition Following,” query and write capabilities for group chat entity fields have been added, including:
- Standard group chat fields
- Custom group chat fields
- Group member-related fields
- Session-related fields
This allows users to make more flexible automated configurations around group chat business needs.
For example:
- Only allow the Agent to take effect in specified types of groups
- Trigger different Skills based on group tags
- Decide whether to reply based on group member identity
- Record customer intent in group chats
- Write business information from group chats into custom fields
This step makes the Main Agent’s reception capabilities in group chats more configurable and better suited for complex team collaboration scenarios.
3. Agent Performance Adds “User Insights”
In v5.2, the Agent performance dashboard adds a User Insights page.
User Insights analyzes customer behavior and consultation content from real session data, helping you understand:
- How many users were reached this week
- Where users primarily come from
- How user consultation trends are changing
- Where users are most likely to drop off
- What questions users ask most frequently
- What characteristics the user profiles show
- Which customer questions are most worthwhile to optimize first
- In which scenarios the Agent can continue to improve
For example, User Insights will display rankings of frequently asked questions, user profile analysis, drop-off nodes in the conversion path, and summaries and optimization suggestions generated based on data.
This allows teams to look beyond message volume and further understand why customers consult, where they hesitate, and which questions affect conversion.
4. Weekly Automatic Sending of Agent Performance Reports
In addition to viewing User Insights in the backend, v5.2 begins to support the weekly sending of Main Agent performance and insight reports.
Currently, the system sends the report every Monday to the email address entered by the space administrator.
The report includes:
- Weekly work summary
- User reach status
- Session and reply performance
- Frequently asked questions by users
- User profile analysis
- Key insights summary
- Recommended directions for priority optimization
For space administrators, even without entering the backend daily, they can understand the overall performance and business changes of the Agent through the weekly report.
In subsequent versions, we will open subscription configuration capabilities, allowing users to customize:
- Push time
- Recipients
- Push method
- Report content scope
IV. Channel Expansion: Support for Integrating WhatsApp Business Accounts via YCloud
In v5.2, 3Chat.ai adds support for integrating WhatsApp Business Accounts via YCloud.
This provides merchants who need to use the WhatsApp Business Platform with a new integration method.
1. What Needs to Be Prepared Before Integration?
Before integrating WhatsApp in 3Chat, you need to prepare a YCloud account and complete the following configurations in YCloud:
- Register or log in to a YCloud account
- Create or select a YCloud Workspace
- Connect or create a WhatsApp Business Account
- Register a WhatsApp business number
- Obtain the YCloud API Key
After completing these preparations, you can return to the WhatsApp integration wizard in 3Chat to continue configuration.
For specific operational references, see: Create a WhatsApp Business API Account | Help Center
2. How to Integrate in 3Chat?
You can find the WhatsApp channel on the Omnichannel page and enter the integration wizard.
You can also enter via “Channels → WhatsApp Business Account.”
The integration process includes:
- Go to YCloud to register or log in
- Complete WhatsApp Business Account and business number configuration in YCloud
- Enter Developer → API Key in the YCloud console
- Copy the default Key or create a new API Key
- Return to 3Chat and fill in the API Key on the WhatsApp integration page
- Click “Update and Sync”
- The system syncs your WABA and WhatsApp business numbers
- View the list of synced numbers and connection status on the page
After integration, the WhatsApp business number will be used as a message channel in 3Chat.
After users send messages via WhatsApp, you can view the messages in the 3Chat inbox, and the 3Chat Agent will participate in reception and replies.
3. Support for Syncing Multiple WhatsApp Business Numbers
If your YCloud space is associated with multiple WhatsApp business numbers, 3Chat will sync the list of available business numbers.
You can view on the integration page:
- WhatsApp business numbers
- Associated WABA
- Current connection status
- Whether integration was successful
For merchants with multiple brands, regions, or business lines, this helps you manage the integration status of different WhatsApp business numbers more clearly.
4. WhatsApp Integration Boundary Description
This WhatsApp integration method is based on YCloud.
Therefore, please note:
- You need to register a YCloud account first
- You need to complete WABA and business number configuration in YCloud first
- 3Chat currently does not complete WABA registration on behalf of users within the platform
- 3Chat completes number synchronization and message integration via the YCloud API Key
- Whether messages can be sent and received normally depends on the status of the business number in YCloud
If the number status is abnormal, you need to handle it first in YCloud or the WhatsApp Business Account side, and then return to 3Chat to sync the status.
V. Summary
3Chat.ai v5.2 is an important upgrade centered around “verifiable, optimizable, and insightful Agents.”
Key updates in this version include:
- Agent Builder officially supports simulation testing
- After the first Agent generation, the Builder automatically recommends core test cases
- Users can directly run recommended cases to automatically execute multi-round simulation tests
- After running recommended cases, they are adopted and accumulated as maintainable test cases
- Users can custom create, edit, copy, and delete test scenarios
- Builder capabilities are further perfected, covering information collection, solution generation, solution modification, knowledge base, real log optimization, simulation testing, and version management
- Knowledge base adds Smart Slicing to improve knowledge retrieval completeness
- Group chat scenario configuration continues to expand, supporting read/write for more group chat entity fields
- Agent performance adds User Insights
- Weekly Agent performance reports are automatically sent to space administrator emails every Monday
- New support for integrating WhatsApp Business Accounts via YCloud
From the launch of Agent Builder in v5.0, to version management in v5.1, to simulation testing and user insights in v5.2, the 3Chat Agent is forming a more complete continuous operations closed-loop.
It not only helps you build an AI customer service agent but also helps you continuously verify, analyze, and optimize this AI Agent, enabling it to more stably participate in sales, service, and customer conversion in real business scenarios.
If you have any questions or suggestions during use, please feel free to post in the community for discussion 
—
The 3Chat.ai Team