Neo4j Aura & Managed Cloud

Neo4j Aura is Neo4j's fully managed cloud service: AuraDB for the database, Aura Graph Analytics for serverless algorithm workloads, and Aura Agent for managed AI agents on top of your graph. Managed does not mean hands-off. Aura removes the operating system, the patching, and the cluster plumbing; it does not remove sizing, cost, data modelling, migration, security design, or the decision of whether it is the right platform for you at all. GraphGuru's Aura consulting covers exactly that layer.

AuraDB sizing and cost optimisation

Aura bills by instance size, and instance size is driven by memory — which in a graph database means the page cache that holds your store. Most overspend we see comes from one of three things: an instance sized for an import that finished a year ago, embeddings stored as LIST<FLOAT> that would take half the space as the native Vector type (2025.10+), or a data model with supernodes and redundant relationships inflating the store.

We measure before we recommend: store size per database, page-cache hit ratio, transaction and query profiles over a representative week. The deliverable is a right-sized instance recommendation with the modelling or storage changes that make it possible, and a projected monthly cost you can hold us to.

Self-managed to Aura migrations

Moving an on-premises or self-hosted Neo4j estate to Aura is a migration like any other, with Aura-specific wrinkles:

• Version alignment. Aura runs current releases; a 4.4 estate must be brought to 5.x compatibility first. Our upgrade & migration practice covers that step.
• Data transfer. Backup-and-restore via the Aura console for moderate sizes; neo4j-admin database import into a fresh instance for very large stores; the Data Importer or Kafka for incremental sync during a parallel-run period.
• Plugins. APOC Core is available on Aura; APOC Extended and custom procedures are not. We inventory your procedure usage and replace what Aura will not run — often with native Cypher that now exists.
• Connectivity. Applications move from bolt:// to the neo4j+s:// routing URI; drivers are upgraded to 6.x where they are not already.
• Parallel run and cutover. The self-managed cluster stays live until sign-off; cutover is a connection-string change.

Aura Graph Analytics: no more ETL for algorithms

Launched in May 2025, Aura Graph Analytics runs Graph Data Science algorithms as serverless sessions against data that can live in AuraDB, in a self-managed Neo4j, or directly in your warehouse — Snowflake, BigQuery, Databricks. For teams whose need is graph features (connected components, PageRank, node embeddings, community detection) rather than a persistent graph application, it eliminates the pipeline that used to copy tables into a graph database every night. Our fraud detection post describes the pattern. We help you decide which workloads belong in Graph Analytics sessions, which belong in a persistent AuraDB, and how to size sessions so algorithm runs stay predictable on cost.

Security, SSO, and network

Aura offers single sign-on, role-based access, private endpoints (VPC/PrivateLink-style peering, depending on cloud), customer-managed encryption keys on the higher tiers, and audit logging. We design the configuration: which roles map to which identity-provider groups, how application credentials are issued and rotated, how private connectivity is set up so the database is never on the public internet, and how backups and logs meet your retention requirements.

Aura Agent

Aura Agent is Neo4j's managed offering for building AI agents that use your graph as their knowledge base. It is in public early access in 2026 with general availability expected later in the year. We are building with it now, and for teams that want a managed GraphRAG path it is worth evaluating — with the caveat that early-access services change. Where the requirements are already firm, we build on neo4j-graphrag and the Neo4j MCP server directly; see our GraphRAG consulting page.

When not to choose Aura

Consultants who only ever recommend the managed service are selling, not advising. Aura is the wrong choice when:

• You depend on plugins it cannot run — APOC Extended, custom procedures, or a GDS workflow that needs a self-managed GDS licence and configuration.
• Data residency or network policy forbids it — a regulated workload that must stay in a specific on-premises environment or a cloud region Aura does not serve.
• Your store is very large and your budget is not — at the top of the size range, self-managed on your own cloud commitment can be materially cheaper, and you may already have the operations team.
• You need configuration Aura does not expose — unusual memory tuning, specific clustering topologies, or operating-system-level integration.
• Your workload is a bursty batch job — spinning up self-managed capacity for a weekly run can beat a permanently provisioned instance.

In those cases we design and operate a self-managed deployment with the same rigour — see our services overview.

How we engage

An Aura assessment is typically one to two weeks: we review your current deployment or plans, model cost, and deliver a written recommendation with a migration or optimisation plan. Implementation follows on your timeline, with senior US-based engineers working inside your team. Contact GraphGuru to start.