Skip to main content
Skip table of contents

Tool-Managed Handover functionality and limitations

Overview

This documentation explains the design limitations and proper configuration of Tool-Managed Handover functionality. Understanding these constraints will help you configure your handover tools correctly and avoid unexpected behaviors.

Key Design Principle

Tool-Managed Handover vs. Fallback Prompts: The system uses either a fallback tool or a fallback prompt, never both. When Tool-Managed Handover is enabled, the configured fallback prompt is bypassed in favor of the tool-based approach.

System Design & Limitations

How Tool-Managed Handover Works

  1. Root Agent Execution: Before breaking out, the root agent executes the configured handover tool

  2. Tool Success: If the tool executes successfully, the handover proceeds as defined by the tool

  3. Tool Failure: If the tool fails for any reason, the system returns a static fallback message that is not configurable or translated into the customer's language.

Current Limitations and Workarounds

  1. Parameter Restrictions: The handover tool cannot have parameters other than those defined in the Template Handover tool, language, and userInput

  2. Workaround for Additional Parameters: If you need other parameters, use User Interactions to collect them before proceeding with the desired functionality.

Group 17.png

Handover tools cannot have other parameters than language and userInput

Group 18.png

Collect parameters using User Interaction

Agent Configuration Constraints

Handover Only Root Agent with Handover Tool - Edge Case

Problem: When a root agent is configured as "Handover Only" and also has the Handover Tool as one of its options, a recursive loop can occur.

Scenario:

  • The user immediately requests, "I want to speak to a live agent."

  • Root agent correctly calls the Handover Tool

  • The tool executes successfully and returns the result to the Root Agent

  • Since Root Agent is "Handover Only", it cannot relay the tool result to the user

  • Root Agent must perform another handover/tool call

  • Root Agent chooses Handover Tool again, asking for the first name repeatedly

Root Cause: Handover Only agents can only perform tool calls or handovers to other agents - they cannot communicate tool results directly back to users.

Solution: Create a dedicated handover agent:

  1. Create New Agent: Design an agent that is not "Handover Only"

  2. Single Purpose: Configure it only to handle calling the Handover Tool or breaking out if the user doesn't want handover

  3. Update Root Agent: Instead of referencing the Handover Tool directly, reference this new Handover Agent

Important Notes

  • When Tool-Managed Handover is active, your custom fallback prompts will not be used

  • Always test your handover tool configuration to ensure it handles edge cases appropriately and does not fail

  • Consider the static fallback message behavior when designing your handover flow

  • Avoid configuring Handover Only agents with direct access to Handover Tools - use dedicated handover agents instead

JavaScript errors detected

Please note, these errors can depend on your browser setup.

If this problem persists, please contact our support.