Building Agents in Neo4j Aura ›Architecture and Components of Aura Agents
The Agent Lifecycle

Building Agents in Neo4j Aura

Architecture and Components of Aura Agents

Design and Implement Simple Aura Agents

Hands-on Challenges

The Agent Lifecycle

Introduction

An Aura Agent does not answer in one step; it receives input, reasons about which tools to use, acts, observes results, and then responds. That sequence is the agent lifecycle.

In this lesson, you will learn the agent’s decision-making sequence: understanding intent, picking tools, executing them, and synthesizing a response.

Steps in the Lifecycle

The agent lifecycle has five steps:

  1. Receive input: User query or message

  2. Reason: LLM plans which tools to use and in what order

  3. Act: Invoke tools: Cypher Template, Similarity Search, or Text2Cypher

  4. Observe: Incorporate tool results into context

  5. Respond: Generate final answer or take next action

Flowchart showing the agent process from user input to LLM reasoning

Why Each Step Matters

  • Receive: The agent must understand user intent before acting

  • Reason: The LLM selects the right tool for the query. For example, a Cypher Template handles "top customers"; Similarity Search finds semantically similar items; Text2Cypher handles ad-hoc questions.

  • Act: The tool executes and returns structured results

  • Observe: The agent adds tool output to its context for the next reasoning step

  • Respond: The LLM synthesizes context into a natural language answer

Trace Through a Query

Walk through an example: "Who has ordered Pavlova repeatedly?"

  1. Receive: The agent receives the question

  2. Reason: The LLM plans which tools to use. The reasoning panel shows the thought process: it identifies that finding repeat Pavlova buyers requires digging into order history and customer details.

    Agent reasoning panel showing detailed thought process and tool selection for finding repeat Pavlova buyers

  3. Act: The agent invokes the Text2Cypher tool, which transforms the natural language query into Cypher and executes it.

    Agent applying Text2Cypher tool showing input query and output with customer names and order counts

  4. Observe: The agent adds the result (customer names and order counts) to context

  5. Respond: The LLM formats a natural language answer listing customers who ordered Pavlova repeatedly

The reasoning panel in the Aura console shows each step, making the agent’s decisions transparent and helping you debug tool selection issues.

Check Your Understanding

Agent Lifecycle

Which step comes after the agent invokes a tool?

  • ❏ Receive input

  • ❏ Reason

  • ✓ Observe: incorporate tool results into context

  • ❏ Respond immediately without observation

Hint

The agent acts, then uses the tool output before deciding the next step.

Solution

Observe: incorporate tool results into context.

After acting, invoking tools, the agent observes the results and incorporates them into its context before generating a response or taking the next action.

Summary

In this lesson, you learned the agent lifecycle: receive, reason, act, observe, respond.

Chatbot

How can I help you today?