Neo4j Aura Agents let you build retrieval systems that query your knowledge graph from the Aura console without writing code.
In this lesson, you will learn:
-
What Neo4j Aura Agents are
-
When to use Aura Agents
-
How Aura Agents work
What are Aura Agents?
A Neo4j Aura Agent is a no/low-code platform in the Aura Console where natural language questions turn into graph queries run against your graph. You connect the agent to an AuraDB instance: your knowledge graph or any graph stored in AuraDB. Ask a question and the agent handles the rest. You build and test agents visually in the Console.
Aura Agents are useful when you want to expose knowledge graph data using natural language, without writing application code.
How Aura Agents work
Aura Agents use reasoning to break down each request into a sequence of steps.
This is often called chain-of-thought or multi-step reasoning: the agent doesn’t jump straight to an answer, it retrieves the information step by step:
-
Interpret user input: Determine what the user needs and which tool or tools to use.
-
Execute tools: Run the selected tool(s) to read data from your graph. The agent may run more than one before answering.
-
Generate response: Use the retrieved data to generate a natural language answer.
Available tools
Aura Agents use three read-only tool types to query your graph.
-
Cypher Template: Runs predefined Cypher queries with parameters you define. This tool is used when you ask questions that can be answered with a fixed query, such as "Get customer ALFKI".
-
Text2Cypher: Generates Cypher from natural language at runtime and executes it against your graph. This tool is used when you ask questions that are not covered by the Cypher Template tools, such as "What are the top 10 products by revenue across all categories?".
-
Similarity Search: Finds nodes by similarity. Use this tool when you ask questions that are not covered by the Cypher Template or Text2Cypher tools, such as "What are the products similar to the product with the name 'Chai'?".
Getting access to Aura Agents
To use Agents, ensure both of the following are enabled in your organization settings:
-
Generative AI assistance
-
Aura Agent. You can toggle this on and off when Generative AI assistance is enabled.
If the agent cannot connect to the instance, check that Tool authentication is enabled for the project. You configure this under Security at the organization level. Tool authentication is enabled by default for organizations created after May 2025. For older organizations, enable it if needed. See Aura Agent documentation for prerequisites.
Before starting the course, check that both toggles are ON. This course uses manual agent creation in the Aura Console: Create with AI and Create from scratch. Agents are under Data Services → Agents in the Aura Console.
To create an agent, you need to be a Project Admin.
Checking your project role
Agent access is controlled by project roles. There are no dedicated agent-level roles.
| Role | Permissions |
|---|---|
Viewer and Member |
Can list and invoke all agents in the project |
Project Admin |
Can also create, update, and delete agents |
To check your role, go to Project → Users in the Aura Console left navigation:
Your role is listed in the Project role column next to your email address:
To give a teammate access to an agent, add them to the Aura project with the appropriate role.
Creating an agent with AI
The Create with AI option generates a working agent from a user-provided prompt and the instance schema.
A prompt is a description of what the agent should do, so the agent can understand the user’s requirements and answer the upcoming question.
For example, if you want to create an agent that answers questions about your customers, you can provide a prompt like this:
You are a customer service agent for Northwind Traders, a food distribution company.
You are to answer questions about customers, orders, products, categories, and suppliers.
You are to decline off-topic or harmful requests.The agent will use the schema of your connected knowledge graph to build tools that query the relevant parts of your graph that match the user’s prompt.
Once you create the agent, you can test it by opening the preview panel and asking it a question, as well as review the reasoning and tool usage.
The preview panel lets you ask the agent questions and see responses. There you can also adjust the agent’s name and description, view the reasoning and tool usage, and edit the prompt to refine the agent’s behavior.
After reviewing and testing the agent, click on Create Agent to save it.
At the end of this process, you will have a working agent that you can use to answer questions about your data using natural language.
Check your understanding
Aura Agent scope
Which of the following is not something an Aura Agent does when it handles a user question?
Select the correct answer.
-
❏ Figure out what the user needs and which tool or tools to use
-
❏ Run the selected tools against the graph (it may run more than one)
-
✓ Write new or updated data back into the graph using tools
-
❏ Use retrieved data to produce a natural language answer
Hint
The lesson describes three steps ending in a natural language answer, and it states that Aura Agents use read-only tool types to query the graph. Writing results into the graph would change stored data.
Solution
Write new or updated data back into the graph using tools.
The agent interprets the request, runs read-only tools to read from the graph, then generates a response. It does not use tools to mutate the graph as part of answering.
Aura Agent tool types
Which of the following are read-only tool types that an Aura Agent can use to query your graph? (Select all that apply)
-
✓ Cypher Template
-
✓ Text2Cypher
-
✓ Similarity Search
-
❏ The Import tool
Hint
The lesson lists three tool types used for querying: parameterized templates, natural language to Cypher at runtime, and similarity-based lookup. Loading or refreshing graph data from files is a separate console workflow, not one of those three agent tools.
Solution
Select Cypher Template, Text2Cypher, and Similarity Search. Those are the three read-only tool types in the lesson. The Import tool is for bringing data into an instance; it is not listed as an Aura Agent tool type.
Summary
In this lesson, you learned what Aura Agents are, how they work, and when to use them.
In the next lesson, you will learn how to:
-
Create an agent with AI
-
Test it and review the output