Plugins in practice
The previous pages describe the mechanics of a plugin — its elements, identities, and communication contracts. This section is about application: concrete scenarios that show how those mechanics combine to solve a real problem. Each use case starts from a business need, states an objective, lists the elements that participate, and walks through the runtime dynamics end-to-end.
Use these pages as blueprints. They deliberately span the two extremes of the model — a rich, multi-element plugin an operator works with live, and a headless, backend-only plugin that reacts to platform events — so you can see how the same contracts serve very different shapes of extension.
| Use case | Elements used | What it demonstrates |
|---|---|---|
| AI Capabilities | Backend only (MCP tools) | Extending the virtual agent's abilities by exposing custom MCP tools — the agent discovers and calls your tools autonomously to read/write your systems, with no widget or settings page. |
| Ice Cream Shop | Backend (MCP + REST) + Widget | Surfacing an internal system (stock, orders) beside a live conversation, for both the virtual agent and the human operator — a unified workspace with no tab-switching. |
| Conversation Observer | Backend only (RTM bridge) | Reacting to platform events in real time: enriching a contact and conversation with tags and metadata from an external database the moment a conversation starts. |