Chat mode transforms the AI Agent action into an interactive experience. Instead of processing a single prompt and moving on, the agent hosts a conversation with an end user until a defined goal is achieved.
❗️Important
Define clear goals in System instructions
The most important part of configuring chat mode is defining a clear, specific goal in your system instructions. The AI uses this goal to determine when the conversation is complete. Without a well-defined goal, the chat may continue indefinitely or end prematurely.
Here's an example of system instructions with a clear goal for chat mode:
"You are a helpful IT support assistant. Use a friendly, concise tone. Your goal is to help users reset their device passwords. Once you've successfully reset the user's password and confirmed they can log in, end the conversation."
Notice the explicit completion criteria: "once you've successfully reset the user's password and confirmed they can log in, end the conversation." This tells the AI exactly when its job is done.
🪄Tip
Initial message
The Initial message field value is the first thing users see when they open the chat. This is where you welcome users and set expectations about what the AI can help with. Unlike the Prompt field in task mode (which provides task instructions to the AI), the initial message is user-facing.
Here's an example:
Hi there! I'm here to help you troubleshoot common IT issues. I can assist with password resets, software installations, and network connectivity problems. What can I help you with today?
🪄Tip
✋ Try this: Use chat mode
Scenario: Employees often have questions when submitting expense reports, such as what receipts are required, which category to use, or how to handle international expenses. You build a chat mode AI Agent action that guides employees through the expense submission process, answers policy questions, and collects the necessary information. Once all required details are gathered and validated, the agent creates the expense report in your system and ends the conversation.
URL identifier and access control
Every agent in chat mode needs a URL identifier. This creates a unique URL where users can access the chat. You can customize this identifier to make it memorable or descriptive.
Access control determines who can access the chat. You have several options:
Public: Anyone with the URL can access the chat.
Secret-based: Users need a secret token to access the chat.
Team members: Only members of specific teams can access.
Tenant members: Only users in your Tines tenant can access.
Choose the access control level that matches your security requirements and use case.
Theme and branding
You can customize the appearance of your chat interface to match your organization's branding. Configuration options include:
Action color: The primary color used for buttons and interactive elements
Page logo: An image displayed at the top of the chat interface
Appearance: Light or dark mode
These options help create a cohesive, professional experience for your users.
Idle timeout
The Idle timeout setting automatically ends the chat after a specified number of minutes without user activity. This helps prevent abandoned chats from consuming resources. The timeout only applies after the user has sent at least one message, and it resets whenever the user starts typing.
A warning is displayed to the user when fewer than 30 seconds remain before the chat ends automatically.
❗️Important
File attachments
Users can attach files to their messages in chat mode. This is incredibly useful for scenarios where users need to share screenshots, error logs, or documents for the AI to analyze.
The supported file types while using the default Tines provider include:
Images:
PNG,JPEG,GIF,WEBPDocuments:
PDF,CSV,MD,TXT,DOC,DOCX,XLS,XLSX
Integrate with Slack
You can connect your agent to Slack, allowing users to interact with the chatbot directly in Slack channels or direct messages. To set this up, navigate to the "Slack app" section in your AI Agent action’s properties panel and follow the on-screen instructions to create and connect a Slack app.
Once connected, users can start conversations with the agent in Slack just like they would with any other bot.