Skip to main content

Centralized AI Handling

LoyJoy lets you run one central AI agent that handles the generative AI for several or all of your agents. You manage the AI configuration in one place, and every connected agent uses it automatically.

info

This guide describes the current approach with the AI Agent module. Older setups used the former AI Gateway, AI Knowledge, and AI Smalltalk modules. These modules are replaced by the AI Agent module; if you still run such a setup, migrate to the AI Agent module first.

1. Create a Central AI Agent​

Create an agent that acts as the central AI. Configure it like a regular agent with an AI Agent module, including its instructions, tools, knowledge, and language settings. This is the configuration that all connected agents will reuse.

To make the central agent reachable for other agents and keep the conversation flowing:

  1. Place the AI Agent module at the start of the process, before the Message start event.
  2. Add a Message start event after the AI Agent module. It contains only an Automatic jump back to the start of the process, so that every incoming customer message is routed into the AI Agent module.
  3. If the Automatic jump cannot point to the AI Agent module directly, add an empty Simple message module at the start of the process and let the jump target that module.

Central AI Agent

2. Connect Other Agents to the Central AI Agent​

In every agent that should use the central AI configuration, do not add an AI Agent module of its own. Instead:

  1. Add a Message start event as usual.
  2. Instead of an AI module, add a Return jump module that jumps to the central AI agent.
  3. After the central agent has processed the message, the flow returns to the original agent automatically.

This way, all connected agents use exactly the same AI configuration, and you only update the AI Agent module in the central agent.

Other Agent

Why Use a Central AI Agent?​

  • Single point of maintenance: update prompts, tools, or knowledge once in the central agent.
  • Consistent behavior: every connected agent answers with the same configuration and tone.
  • Clear ownership: one agent owns the AI logic, other agents only route to it.