GenAI resource types

This section describes each GenAI resource type that can be included in a local package, what is exported from the source tenant, and how it behaves on the target tenant after deployment.

Tools

A tool is a configurable action that an agent can invoke, such as calling a REST API or running a query. Tools connect an agent to the applications and data it acts on.

Source tenant

  • All enabled and disabled tools are available for selection.
  • Tools that invoke custom API suites are included. You must manually add the API suite reference for each tool.
  • MCP Tools are included when an agent or skill in the package references the MCP Tool.

Target tenant

Behavior Detail
Deployed as Retains its enabled or disabled status from the source tenant.
Editability Fully editable. No restrictions.
Created By Set to System.
Duplicates If a tool with the same name already exists on the target, import validation reports an error. The deployment does not create duplicates.

Agents

An agent is a domain-specific assistant that interprets your request and decides which tools and skills to use to fulfill it.

Source tenant

  • All enabled and disabled agents are available for selection.
  • You must manually add the custom security roles (SRs) and functional security roles (FSRs) assigned to the agent.
  • You must manually add the tools and skills referenced by the agent.
  • MCP Tools assigned to the agent are included.

Target tenant

Behavior Detail
Deployed as Retains its enabled or disabled status from the source tenant.
Editability Fully editable. No restrictions.
Created By Set to System.
Duplicates If an agent with the same name already exists on the target, import validation reports an error. The deployment does not create duplicates.
API Gateway The Created By field for imported agents is set to System in the API Gateway.

Skills

A skill is a reusable bundle that contains context and optional tool bindings. A skill extends an agent's instructions during runtime.

Source tenant

  • All enabled and disabled agents are available for selection.
  • You must manually add the tools referenced by the skill.
  • You must manually add the custom security roles assigned to the skill.

Target tenant

Behavior Detail
Deployed as Retains its enabled or disabled status from the source tenant.
Editability Fully editable. No restrictions.
Created By Set to System.
Duplicates If a skill with the same name already exists on the target, import validation reports an error.

Extensions

An extension injects an instruction block into a target agent or skill when specified conditions are met.

Source tenant

  • All enabled and disabled agents are available for selection.
  • You must manually add the agents and skills the extension targets.

Target tenant

Behavior Detail
Deployed as Retains its enabled or disabled status from the source tenant.
Editability Fully editable. No restrictions.
Created By Set to System.
Duplicates If an extension with the same name already exists on the target, import validation reports an error.

Focus Modes

A focus mode limits the GenAI Assistant to a curated set of agents. You see agents that are relevant to a specific task or role.

Source tenant

  • All enabled and disabled focus modes are available for selection.
  • You must manually add the agents and tools referenced by the focus mode.

Target tenant

Behavior Detail
Deployed as Retains its enabled or disabled status from the source tenant.
Editability Fully editable. No restrictions.
Created By Set to System.
Duplicates If a focus mode with the same name already exists on the target, import validation reports an error.

Guardrails

A guardrail is an administrator-defined rule that blocks specified topics or content in prompts and responses.

Source tenant

  • All enabled and disabled guardrails are available for selection.

Target tenant

Behavior Detail
Deployed as Retains its enabled or disabled status from the source tenant.
Editability Fully editable. No restrictions.
Created By Set to System.
Duplicates If a guardrail with the same name already exists on the target, import validation reports an error.

Conversation Starters

A conversation starter is a preconfigured prompt suggestion that is displayed in the GenAI Assistant. A conversation starter helps you begin a common task.

Source tenant

  • All conversation starters are available for selection.
  • Conversation starters have no dependencies on other GenAI content types.

Target tenant

Behavior Detail
Editability Fully editable. No restrictions.
Created By Set to System.
Duplicates Conversation starters can create duplicate conversation starters on the target environment when the same conversation starter already exists. Review duplicate conversation starters and remove the duplicates manually after deployment.
Note: Unlike other GenAI content types, conversation starters can create duplicate entries on the target tenant when an identical conversation starter already exists. After deployment, review the Conversation Starters list and remove duplicate entries.

Prompts (Prompt Catalog)

A prompt is a Prompt Catalog entry that invokes a model directly through a saved instruction. A prompt does not use an agent.

Source tenant

  • Only prompts in Published status can be included in the export..
  • Prompt settings are included. Prompt settings include the model, the temperature, and the system prompt.
  • Attachments and variables that are configured on the prompt are not included.

Target tenant

Behavior Detail
Deployed as Retains its status (Published) from the source tenant.
Editability Fully editable. No restrictions.
Created By Set to System.
Attachments and Variables Not transferred. You must re-create the attachments and variables manually on the target tenant.
Duplicates If a prompt with the same name already exists on the target, import validation reports an error.