The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →To create an AI agent with Neo4j Aura Agent, enable Generative AI assistance and Aura Agent for your organization, turn on Tool authentication for the project, and connect the agent to a knowledge graph in AuraDB. In the Aura console, choose Agents > Create Agent, configure the agent manually or draft it with AI, add retrieval tools, test its answers, and save. Keep it internal while developing; making an agent externally available through REST or MCP incurs charges.
What you need before creating an agent
Aura Agent builds GraphRAG agents that retrieve context from a knowledge graph in AuraDB. Neo4j describes it as “a no/low-code agent platform that allows you to build, test, and deploy GraphRAG agents contextualized by your own knowledge graph in AuraDB.” Start with an Aura account and a graph in an AuraDB instance, then check that the organization and project settings permit agent use.
- In organization settings, enable Generative AI assistance and Aura Agent.
- In organization security settings, enable Tool authentication for the project.
- Confirm you have the appropriate project role. Project admins can create, edit, and delete agents; members and viewers can list and use agents.
- Have a running instance ready if you plan to test the agent or use AI-assisted creation. Manual configuration can be set up while the instance is not running, but testing requires it to run.
Neo4j’s official tutorial estimates 30–45 minutes to complete its example; that is a tutorial estimate, not a guarantee of the time required to build and validate an agent for your own graph. See the Aura Agent documentation and Neo4j Developer Guides tutorial for current setup details.
Choose how to create the agent
In the Aura console, open Agents and select Create Agent. Choose the approach that suits how much control or drafting help you want:
Recommended Free Tools
#1 Best Overall
| Approach | What you configure | Instance requirement | Important trade-off |
|---|---|---|---|
| Manual creation | Select the instance, provide a name and description, add optional prompt instructions, and choose tools. | The instance need not be running to configure the agent; it must run for testing. | Gives you direct control over instructions and retrieval tools. |
| Create with AI | Select an instance and write a detailed prompt describing the domain, audience, tasks, and example questions. Review the generated description, instructions, and tools. | The instance must be running. | AI drafts the configuration from your prompt and database schema. Regenerate with AI overwrites the current configuration. |
AI generation is a starting point, not validation. Review the generated tools and instructions against your graph’s schema and the questions users will ask. If you configure Similarity Search, also confirm that the graph already has text embeddings and select an embedding provider and model compatible with those stored vectors.
Select retrieval tools that fit your graph and questions
An agent’s usefulness depends on whether its retrieval tools can answer the questions you expect. You can configure the tools manually or review those suggested during AI-assisted creation.
Rank #2
| Tool | Best fit | Requirements and setup | Key consideration |
|---|---|---|---|
| Cypher Template | Repeated, predictable questions, complex queries, or well-defined business rules. | Define parameter names, types, and descriptions, then write and test the query. | Return only relevant properties rather than duplicate data, embeddings, or full graph elements. Limit results to about 10–50 rows where appropriate, as Neo4j advises. |
| Similarity Search | Semantic matching, document search, content discovery, or finding similar clauses and terms. | Requires text embeddings and a vector index. Choose the index and Top K; optionally add a Cypher post-processing query to retrieve connected graph context. | Use the same embedding model as the stored vectors. Confirm the live console’s model options before creating or configuring an index. |
| Text2Cypher | Dynamic question-to-query retrieval when a fixed template or similarity search is not a good fit. | The tool uses the question, database schema, and its system prompt to generate a Cypher query. | Explain when to use it, domain-specific schema details and identifiers, and appropriate aggregations so the generated query has useful context. |
Check embedding model names carefully
Neo4j’s current model disclosure lists Google embeddings gemini-embedding-001, text-embedding-005, and text-multilingual-embedding-002; it also lists Azure OpenAI embeddings text-embedding-3-small, text-embedding-3-large, and text-embedding-002. The Aura Agent Similarity Search documentation also names text-embedding-ada-002. Since the names in these official references differ, do not assume they are interchangeable: verify the available options and current model disclosure before configuring the index. See Aura Agent documentation and Neo4j’s AI model disclosure.
Build and test the agent in the Aura console
- Open Agents in the Aura console and click Create Agent.
- Choose manual creation or Create with AI. Select the target AuraDB instance and provide a clear name and description.
- Add prompt instructions that define the agent’s scope and how it should respond. For AI-assisted creation, give a detailed prompt with the domain, audience, tasks, and sample questions.
- Add or review retrieval tools. Configure their parameters and descriptions to match the graph and the questions users will ask.
- Ask representative questions and inspect the tool sequence and results. If the agent selects the wrong tool, refine the tool descriptions or instructions and test again.
- Save after the behavior is satisfactory. Keep the agent internal while you are still evaluating it.
Neo4j’s tutorial examples include “How many Python developers do I have?”, “Who is most similar to Lucas Martinez?”, and “Which individuals have collaborated to deliver the most AI Things?” These refer to the tutorial graph; they do not imply that those entities or answers exist in your database.
Rank #3
Use structured evaluations for repeatable checks
For a more consistent test, create an evaluation dataset containing test questions and expected answers, with expected tool calls if useful. Neo4j supports reusable datasets within an Aura project, with up to 50 questions. Run evaluations before promoting an agent and again after changing its prompt or tools. Treat evaluation scores as diagnostic signals: they can reveal regressions or mismatches, but they are not independent proof that answers are correct.
Decide whether the agent should be internal or external
Internal agents are available for use within an Aura project and are free to use according to Neo4j’s documentation. External access can expose an agent through REST or MCP and incurs charges. Check Neo4j’s current pricing and Aura billing documentation before estimating costs; an exact current price is not established here.
Rank #4
| Access method | How access works | What to consider |
|---|---|---|
| REST | Make the agent externally available, copy its endpoint, obtain a bearer token using Aura API client credentials, and send the user’s question to the agent endpoint. The response is structured JSON. | Protect the API credentials and account for external-use charges. |
| MCP | Enable the MCP server for the externally available agent. Clients can use user authorization or machine-to-machine client credentials. | Neo4j documents a limit of 15 requests per hour per client ID for its MCP token endpoint and recommends caching a token for its full expiration period. MCP exposes Aura Agent as a read-only server to an external client. |
Account for data location and model governance
Neo4j states that all Aura Agents are hosted in Belgium in GCP europe-west1 and that all interactions go via Belgium. Neo4j selects the models centrally; users cannot change them, and the models may be updated. Confirm that these arrangements meet your organization’s data-residency and governance requirements before using the service. See Neo4j Aura Agent documentation.
Quick Recap
Common setup issues to check
- You cannot create or use an agent: verify that Generative AI assistance and Aura Agent are enabled for the organization and Tool authentication is enabled for the project.
- AI creation or testing is unavailable: make sure the selected AuraDB instance is running. Manual setup does not require a running instance, but testing does.
- Similarity Search returns poor matches or is unavailable: check that embeddings and a vector index exist, and that the selected model matches the one used to create the stored vectors.
- The agent uses the wrong retrieval tool: revise tool descriptions and prompt instructions, then retest with representative questions and inspect the tool sequence.
- A regeneration changes your setup: Regenerate with AI overwrites the existing agent configuration, so review the result before relying on it.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
