Capabilities

Learn how to enable, configure, and manage the built-in tools and capabilities your AI agent can use—such as sending files, answering FAQs, collecting user information, and escalating to human agents—across supported channels.

Manage your AI agent’s capabilities through two types of tools: predefined tools that come built-in with the platform, and custom tools that you can create to extend your agent’s functionality.

Tools (Predefined)

The Tools (Predefined) tab allows you to enable or disable built-in capabilities that your AI agent can use during conversations. Each tool represents a predefined function—such as answering FAQs, sending files, or escalating to a human agent—that can be toggled on or off depending on your business needs.

When a tool is enabled, additional configuration may be required before the agent can use it successfully. Some tools also depend on the type of channel where the end user is interacting (e.g., WhatsApp, Zendesk, Freshchat).

---

How to Enable or Disable a Tool

1. Navigate to Agents > Capabilities in the left menu.
2. Under Tools (Predefined), locate the tool you want to manage.
3. Use the toggle switch to turn the tool on or off.
4. If the tool requires configuration, click Configuration, complete the required setup, and save your changes.
5. Check that the tool is supported by the channels your agent is connected to.

---

Available Predefined Tools

Send Files

• Description: Allows the agent to send files directly to the user during a conversation.
• Default: Always enabled.
• Configuration: No setup required.

---

Answer FAQs

• Description: Enables the agent to respond to frequently asked questions about your product or service.
• Default: Always enabled.
• Configuration: No setup required. You can manage and update FAQs in the Knowledge section of the agent.

---

Collect Information

• Description: Enables the agent to collect, verify, and store user information (e.g., name, email, phone number) for lead generation or service requests.
• Toggle: Off by default.
• Configuration: Define the fields you want the agent to collect and where the data should be stored.

This tool is so important that once enabled, it activates a dedicated Collect Information tab in the Capabilities section. From there, you can create, edit, and manage information collection configurations in detail.

How to Configure Collect Information

1. Go to Agents > Capabilities > Collect Information.
2. Click New Configuration to start.
3. Enter a Name for your configuration (e.g., registro_interesados).
4. Add any Special Instructions to guide the agent (e.g., always capture user details if they show interest in your product).
5. Define the Fields to Collect (e.g., name, surname, email, phone). Mark fields as required if necessary.
6. Choose the Integration Channels where this configuration will apply (e.g., WhatsApp, Semantiks Console, Freshchat).
7. (Optional) Add Trigger Topics to specify keywords or intents that activate the collection process.
8. Save by clicking Update configuration.

Once active, your configuration will appear in the Collect Information tab, showing:

• Fields to collect
• Trigger topics
• Associated channels
• Status (Active or Inactive)

---

4. Human Escalation (Channel-Dependent)

• Description: Allows the agent to transfer a user from an AI conversation to a live support agent.
• Toggle: Off by default.
• Channel Dependency:
  • This tool only appears when the agent is connected to certain customer service platforms such as Zendesk or Freshchat.

How to Configure Human Escalation

1. Connect your Zendesk or Freshchat account.
2. Navigate to Agents > Capabilities > Tools (Predefined).
3. Locate Human Escalation and toggle it on.
4. Define escalation rules (e.g., trigger phrases, fallback scenarios).
5. Save your configuration to activate.

---

Key Notes

• Tools marked Default (Send Files, Answer FAQs) are always available and cannot be disabled.
• Tools that require configuration must be set up before they can be used by the agent.
• Some tools (e.g., Human Escalation) will only appear if the agent is connected to a compatible channel.
• Collect Information is unique: enabling it adds its own tab for deeper configuration.
• You can revisit this tab at any time to adjust your enabled capabilities.

Tools (Custom)

The Tools (Custom) tab allows your organization to extend the AI agent’s functionality by creating custom tools that integrate with your internal systems or third-party APIs. Unlike predefined tools, these are designed and managed entirely by your organization, offering flexibility to meet unique business needs.

You can create tools for tasks such as querying proprietary databases, invoking external APIs, or applying custom business logic. Once created, these tools can be invoked by the agent during conversations.

How to Create a Custom Tool

1. Navigate to Agents > Capabilities > Tools (Custom).
2. Click Create Tool.
3. Fill in the required fields in the tool creation form:

General Information

• Name: Enter a unique tool name.
• Description: Describe what the tool does for documentation and clarity.
• HTTP Method: Select the request method (GET, POST, etc.).
• Timeout: Define the maximum time (in seconds) the agent should wait for a response.
• URL: Provide the endpoint the tool will call (e.g., https://api.example.com/endpoint).
• Visibility: Tools are currently limited to Private within your organization.

Tool Properties

• Add key-value properties that may be required for execution.
• For example, an API key or parameter value.

Prompt and Output

• Prompt Description: Explain how the agent should use the tool and what format the response should follow.
• Output Example: Provide a sample response for validation (e.g., JSON structure).
• Schema Fields: Use the schema builder or JSON editor to specify the expected output structure.
• Output Validation: Optionally enable validation and configure how errors should be handled (Warn, Error).

Error Handling

• Add logic or instructions on how the agent should react to errors returned by this tool.
• Example: If API call fails, return a fallback message such as “Sorry, this service is not available right now.”

4. Click Save to create the tool.

Tool Invocations

Once tools are created, the Tool Invocations sub-tab allows you to monitor how custom tools are being called by the agent:

• View logs of past invocations
• Inspect request and response details
• Troubleshoot errors or unexpected behaviors

Key Notes

• Custom tools are organization-scoped, meaning they can be reused across multiple agents in your account
• At present, all tools are Private to your organization; public or shared tools are not supported
• Well-defined output schemas improve reliability by ensuring the agent understands the tool’s responses
• Always provide error handling logic so the agent can gracefully manage failures

Example: Creating a Custom Weather API Tool

Step 1: Name: Weather Lookup
Step 2: Description: Fetches current weather data for a given city
Step 3: HTTP Method: GET
Step 4: Timeout: 10 seconds
Step 5: URL: https://api.weather.com/v3/weather?city={city}
Step 6: Tool Properties: Add a property for city (string)
Step 7: Prompt Description: Use this tool to return the current weather when a user asks about weather conditions
Step 8: Output Example:

{
  "temperature": "28C",
  "condition": "Sunny"
}

Step 9: Schema Fields:

• temperature (string, required)
• condition (string, required)

Step 10: Error Handling: If API call fails, return a fallback message: “Sorry, I can’t fetch the weather right now.”