Manager Agent in Lyzr Agents

Overview

The Manager Agent architecture in Lyzr allows for modular, composable workflows where one agent (the manager) can coordinate and delegate tasks to multiple other specialized agents (called managed agents). This enables powerful, flexible automation where complex multi-step operations can be executed efficiently by splitting responsibilities between agents.

For example, in a fitness assistant setup:

• The main agent provides personalized fitness advice.
• A managed agent like Mail Generator can take the generated diet plan or fitness routine and email it to the user.

---

How It Works

Structure in Agent JSON

In the agent JSON configuration, the managed_agents field is used to declare and reference one or more agents that the main agent can call to perform specific sub-tasks.

"managed_agents": [
  {
    "id": "67f632f43e4bcae4a5cfd109",
    "name": "Mail Generator",
    "usage_description": "Generate a proper diet plan and mail it to the user, whenever the user starts to chat ask for their mail ID 1st"
  }
]

Fields Breakdown

Field	Type	Description
id	string	Unique identifier of the managed agent (retrieved from the Lyzr Agent database).
name	string	Human-readable name of the agent for referencing and understanding purpose.
usage_description	string	Description of how and when to invoke this agent. Acts as a system-level guide to help orchestrate agent responsibilities.

---

Benefits of Manager Agent Structure

• Separation of Concerns: Each agent handles a specific role (e.g., fitness vs. email).
• Reusability: Managed agents like Mail Generator can be reused across multiple parent agents.
• Scalability: Adding new capabilities only requires connecting additional managed agents without modifying the main logic.
• Workflow Automation: Enables seamless multi-step workflows coordinated by the manager agent.

---

Example Use Case

Scenario

A user interacts with a Fitness Assistant agent which:

1. Asks for their fitness goals and suggests a workout plan.
2. Triggers a Mail Generator agent to email a corresponding diet plan.

JSON Snippet

{
  "name": "Trial",
  "description": "Fitness assistant",
  ...
  "managed_agents": [
    {
      "id": "67f632f43e4bcae4a5cfd109",
      "name": "Mail Generator",
      "usage_description": "Generate a proper diet plan and mail it to the user, whenever the user starts to chat ask for their mail ID 1st"
    }
  ]
}

---

Implementation Notes

• Manager agents can dynamically call the managed_agents based on internal logic, conditions, or user inputs.
• All agents must be registered within the Lyzr ecosystem with a valid agent id.
• Ensure usage_description is descriptive enough to automate orchestration decisions correctly.
• Agent calling sequence can be rule-based or LLM-driven depending on the parent agent’s configuration.

---

Best Practices

• Keep agents single-responsibility for maintainability.
• Reuse agents where possible to reduce redundancy.
• Add proper validation when passing data between agents.
• Define fallback mechanisms in the manager in case a managed agent fails.

---

Summary

The managed_agents feature of Lyzr empowers developers and builders to create composable, intelligent agent ecosystems where each agent can focus on its strength. With manager agents coordinating the workflow, complex tasks like fitness guidance followed by email distribution become seamless and modular.