Introduction
Aura Agents let you build GraphRAG-style agents in the Aura console without writing code. Understanding what they are and when to use them helps you choose the right approach for your application.
In this lesson, you will learn what Aura Agents are, when to use them, and how they fit into the Neo4j ecosystem.
What are Aura Agents?
A Neo4j Aura Agent is a no/low-code agent platform built into the Aura console. It is backed by a knowledge graph stored in AuraDB, interpreting user questions and invoking the right tools to retrieve answers.
The agent interprets your question, finds relevant information in the graph, and invokes the appropriate tool: Similarity Search, Text2Cypher, or Cypher Template.
An Aura Agent consists of:
-
LLM for planning, reasoning, and tool selection
-
Tools for querying the knowledge graph: Cypher Template, Similarity Search, and Text2Cypher (read-only)
-
Knowledge graph — your data stored in AuraDB
-
Orchestrator to coordinate the workflow
Conversation history is not currently persisted between sessions. For the full reference, see Aura Agent documentation.
Prerequisites and Pricing
To create an Aura Agent:
-
Have a knowledge graph in a running AuraDB instance
-
Enable Generative AI assistance in your Organization settings
-
Toggle on Aura Agent (available when Generative AI assistance is enabled)
Keep Generative AI assistance enabled
Disabling Generative AI assistance automatically disables Aura Agent.
Visibility: Your agent has two visibility options — Internal or External.
-
Internal (default, free): You can create the agent, add tools, and test it within the Aura console. Only members of your Aura project can use it.
-
External (incurs charges): The agent is exposed via an Aura API endpoint and/or an MCP server endpoint, so you can integrate it into your AI application or connect it to an MCP client such as Claude Desktop or Cursor. See Neo4j pricing for details.
Location: All Aura Agents run in Belgium (GCP europe-west1). All agent interactions route through this region, including internal testing in the console.
AI models: Aura Agent uses a curated set of AI models for planning, reasoning, retrieval, and explainability. See How Aura Agent uses AI models.
Agentic Concepts
-
Agents: Systems that perceive, reason, and act to answer questions grounded in your knowledge graph
-
Tools: Capabilities the agent invokes: Cypher Template, Similarity Search, and Text2Cypher
-
Retrieval: GraphRAG and vector search to fetch relevant context from your AuraDB graph
Aura Agents are useful when:
-
Users ask natural language questions over complex, connected data in AuraDB
-
Queries require multi-hop reasoning across relationships in the graph
-
You want to expose graph knowledge to an external application or MCP client without writing code
Two paths to Neo4j agents
This course uses Aura Agents, a no/low-code visual approach to create, configure, and test agents directly in the Aura console. For a code-based approach with Python and LangChain, take the Neo4j and Generative AI Workshop.
Check Your Understanding
GraphRAG-style Agents
When are GraphRAG-style Agents most useful?
-
❏ Only for simple key-value lookups
-
✓ When queries require multi-hop reasoning across your graph
-
❏ When you need to run batch imports
-
❏ Only for administrative tasks
Hint
Agents reason over connected data and can invoke tools. They excel when the task involves traversal and reasoning.
Solution
When queries require multi-hop reasoning across your graph.
Agents are useful for multi-hop reasoning, natural language questions over complex connected data, and when your application combines graph results with other data or APIs in your own code.
Summary
In this lesson, you learned what Aura Agents are, their components, visibility options, and when to use them.