Conversational Platform
LoyJoy in a consolidated AI stack
Why this is not a second AI tool competing with your assistant decision.
Two decisions, routinely merged into one
- Which assistant do employees work in?
- Copilot, ChatGPT Enterprise, Claude, Gemini. An identity, licensing and shadow-AI question. Consolidating it is sound.
- Where are customer-facing agents built, operated and evidenced?
- A process, integration and audit question, touching end-customer personal data, telephony, product data and complaint handling.
LoyJoy answers the second and is deliberately agnostic about the first. Standardising the assistant does not require replacing the layer the agents run in, and keeping LoyJoy does not constrain the assistant decision.
Model Context Protocol, in both directions
| Direction | Effect |
|---|---|
| LoyJoy as MCP client | A LoyJoy agent uses your systems as tools through their MCP servers or APIs: ERP, shop, CRM. Configured, not coded. |
| Agent as MCP server | External AI assistants reach your services and processes through your agent, so the customer interface stays yours as consumer assistants start acting for customers. |
| Tenant as MCP server | Your teams work with analytics, process definitions and knowledge from inside the approved assistant environment. Setup as a custom connector takes minutes. |
Whichever assistant the group standardises on, the integration path is the same protocol. Model choice stays open as well: the default is Gemma 4 on LoyJoy hardware, higher tiers add GPT, Claude and Mistral, and your own Azure key is supported. Neither decision has to wait for the other.