Design Aura Agents

Introduction

Prompt design drives how well the agent chooses tools and stays in scope. Clear role, scope, and tool descriptions reduce wrong tool selection and off-topic answers.

In this lesson, you will learn how to write effective prompts and configure agents for predictable behavior.

What Makes a Good Agent Prompt

An agent prompt tells the LLM who it is and what it should do. A complete prompt includes:

Role: What the agent represents, such as "retail analyst for Northwind"
Scope: What questions it can answer and what is out of bounds
Tools: Which tools are available and when to use each one
Examples (optional): Sample question-answer pairs for complex tasks

Best Practices

Keep the scope focused. One main task per agent works better than a do-everything agent. If you need different capabilities, create multiple agents.

Write clear tool descriptions. The LLM picks tools based on their descriptions. If a description is vague, the agent may select the wrong tool.

Include the graph schema in your prompt. Mentioning node labels and relationship types helps the LLM construct valid queries.

Test edge cases. Try ambiguous questions, queries that return no results, and off-topic requests.

Add safeguards. Include instructions for handling harmful or irrelevant queries.

Example: Northwind Retail Analyst

Here is a sample configuration for a Northwind agent:

Element	Value
Goal	Answer questions about customers, orders, products, and sales patterns
Tools	Get Customer, Top Customers by Order Count, Products by Category
Prompt	"You are a Northwind retail analyst. You have access to Cypher tools to query the graph. Answer questions about customers, orders, products, and sales."
Safeguard	"If the question is off-topic or asks for harmful content, politely decline."

Element

Value

Goal

Answer questions about customers, orders, products, and sales patterns

Tools

Get Customer, Top Customers by Order Count, Products by Category

Prompt

"You are a Northwind retail analyst. You have access to Cypher tools to query the graph. Answer questions about customers, orders, products, and sales."

Safeguard

"If the question is off-topic or asks for harmful content, politely decline."

Once you save the agent, it displays a chat interface. Users see the description and can start asking questions.

Agent chat interface showing the Online Sales Assistant with description and Ask anything prompt

From Schema and Use Case to Agent

mermaid

From schema, use case, prompts, and tools to a ready-to-deploy agent

%%{init: {
  "theme": "base",
  "securityLevel": "strict",
  "fontFamily": "Public Sans, Arial, Helvetica, sans-serif",
  "themeVariables": {
    "background": "#F5F7FA",
    "primaryTextColor": "#014063",
    "fontSize": "14px",
    "primaryColor": "#eef6f9",
    "primaryBorderColor": "#014063",
    "lineColor": "#485662"
  }
}}%%
flowchart LR
    SCHEMA["Graph Schema"]
    USECASE["Use Case Description"]

    PROMPTS["Prompts"]
    TOOLS["Tools"]
    AGENT["Ready to Deploy Agent"]

    SCHEMA -->|feeds into| TOOLS
    USECASE -->|feeds into| PROMPTS
    USECASE -->|feeds into| TOOLS
    PROMPTS --> AGENT
    TOOLS --> AGENT

    style SCHEMA fill:#edf6e8,stroke:#006E58
    style USECASE fill:#eef6f9,stroke:#014063
    style PROMPTS fill:#E7FAFB,stroke:#006E58
    style TOOLS fill:#E7FAFB,stroke:#006E58
    style AGENT fill:#014063,stroke:#014063,color:#fff

The graph schema informs your tools: Cypher patterns, node labels, relationships. The use case description drives both prompts (role, scope, examples) and tool selection. Prompts and tools together produce a ready-to-deploy agent. See Aura Agent documentation for prompt and tool configuration details.

Check Your Understanding

Designing Agent Prompts

Which of the following is a best practice when designing an Aura Agent prompt?

❏ Use a single long prompt that covers every possible scenario
✓ Define the agent’s role and scope clearly, and specify available tools
❏ Avoid mentioning the graph schema in the prompt
❏ Never include examples to keep the prompt short

Hint

Effective prompts give the agent a clear purpose and context. Specifying tools helps the LLM choose correctly.

Solution

Define the agent’s role and scope clearly, and specify available tools.

mermaid

%%{init: {
  "theme": "base",
  "securityLevel": "strict",
  "fontFamily": "Public Sans, Arial, Helvetica, sans-serif",
  "themeVariables": {
    "background": "#F5F7FA",
    "primaryTextColor": "#014063",
    "fontSize": "14px",
    "primaryColor": "#eef6f9",
    "primaryBorderColor": "#014063",
    "lineColor": "#485662"
  }
}}%%
flowchart LR
    ROLE["Clear Role"]
    SCOPE["Clear Scope"]
    TOOLS["Tool Descriptions"]

    PROMPT["Effective Prompt"]
    LLM["LLM Reasons"]
    INVOKE["Correct Tool Invocation"]

    ROLE --> PROMPT
    SCOPE --> PROMPT
    TOOLS --> PROMPT
    PROMPT --> LLM
    LLM --> INVOKE

    style ROLE fill:#edf6e8,stroke:#006E58
    style SCOPE fill:#edf6e8,stroke:#006E58
    style TOOLS fill:#edf6e8,stroke:#006E58
    style PROMPT fill:#eef6f9,stroke:#014063
    style LLM fill:#E7FAFB,stroke:#006E58
    style INVOKE fill:#014063,stroke:#014063,color:#fff

Clear role and scope plus tool descriptions help the LLM reason about when to invoke each tool.

Summary

In this lesson, you learned how to design agents with clear prompts, focused scope, and effective tool descriptions.

Building Agents in Neo4j Aura

Architecture and Components of Aura Agents

Design and Implement Simple Aura Agents

Hands-on Challenges

Design Aura Agents

Introduction

What Makes a Good Agent Prompt

Best Practices

Example: Northwind Retail Analyst

From Schema and Use Case to Agent

Check Your Understanding

Designing Agent Prompts

Summary

Chatbot