Skip to main content

How to deploy Agentforce Subagent AKA Topic (GenAiPlugin) metadata

A walkthrough on how to deploy Agentforce Subagent (Topic) metadata

Written by Gareth Hefferon

Important to note: Make sure to use at least API version 60 for the comparison to retrieve the Agentforce Subagent (Topic) metadata type.

Gearset's Agentforce metadata deployment solution simplifies the process of deploying AI-powered configurations in Salesforce, including the Agentforce subagent (Topic) metadata. Agentforce Subagents determine the scope that an AI Agent has i.e. Customer service, sales, etc.
​
This guide will explain how to successfully deploy the Subagent (GenAiPlugin) metadata with Gearset.

Note: When a custom subagent (topic) is created based on a default subagent, it becomes embedded as metadata within the GenAiPlannerBundle and will not appear as a separate item in the comparison. Please check the agent planner metadata to view your subagent.

Prerequisites

Make sure to use at least API version 60 for the comparison. Read the prerequisite section of the How to deploy GenAi metadata guide to find out how to do this.

What metadata is needed in the filter?

In your metadata filter you will only need to include GenAiPlugin.
​

How to deploy GenAiPlugin metadata through Gearset

To deploy a GenAiPlugin metadata type, you need to make sure that all the GenAiFunction dependencies exist on the target. If they don't then you need to deploy them alongside the GenAiPlugin for the validation/deployment to be successful.

Example:

This GenAiPlugin is dependent on the following GenAiFunctions:

  • Check_Weather

  • Generate_Personalized_Schedule

  • EmployeeCopilot_GetRecordDetails

  • Issue_Bulk_Resort_Credits

  • Issue_Resort_Credit

  • EmployeeCopilot__QueryRecords

  • EmployeeCopilot_SummarizeRecord

If any of these GenAiFunctions are not in the target org, or being deployed to the target org along with the GenAiPlugin, then the validation/deployment will fail.
​
We will be adding more problem analyzers over time to help highlight any missing GenAiFunction dependencies before a validation or deployment occurs.

Did this answer your question?