From the NetApp edition of October 31, 2026
NetApp declares AI Data Engine 1.3 its first generally available release, with metadata cataloging, lifecycle actions and governed MCP access
NetApp · Strategic partner: AI data services and metadata management · Global
- Publisher
- NetApp Docs (AI Data Engine)
- Original publication
- Release entry dated 26 October 2026; page last updated 10/30/2026
- Source retrieved
- 2026-11-01
- Event date
- 2026-10-26
What happened
NetApp's AI Data Engine (AIDE) documentation states that version 1.3, dated 26 October 2026, is the first generally available release. AIDE is described as a metadata intelligence service managed through NetApp Console. It scans connected file shares, database servers and StorageGRID systems into a metadata catalog. It flags cold, stale and duplicate data, open-permission directories and oversized or unused database tables. It can copy or archive files to NFS or SMB targets, tag them or permanently delete them, and save filters as monitored policies. It offers governed MCP access for Claude Code, Claude Desktop, Cursor and OpenCode, with Microsoft Entra ID scopes per collection. The 1.3 notes do not mention vectorization, semantic enrichment, Data Curator or Data Guardrails, which StorageMath, HyperFRAME and Futurum described as part of the earlier AIDE positioning.
Why it matters
SLED institutions with large unstructured file estates, such as county file servers, university research shares and district home directories, can now evaluate AIDE as a supported product rather than an early-access program. The functions that ship are data hygiene and governance: finding stale and duplicate data, open permissions and retention-driven deletion. These map to records-retention and storage-cost work more directly than to generative AI. There is no SLED deployment evidence yet.
Evidence and measured results
Vendor release notes only. Timeline context from independent analysts: Futurum (March 2026) reported management expected GA in Q4 FY2026, and HyperFRAME (19 March 2026) reported broader rollout expected in early summer. GA landed on 26 October, months later than both. The companion deployment overview (10/23/2026) gives sizing baselines of up to about 3 billion files single-node and 6 billion multi-node, which NetApp calls recommended baselines rather than hard limits. No benchmarks, customer counts or accuracy measures are given.
Limitations and uncertainty
Vendor-authored. GA status does not establish production scale, accuracy or cost outcomes. The 1.3 feature list is narrower than the 2025 announcement, and this run did not confirm whether the missing features are deferred or dropped. Known limitations were covered in our 2026-10-29 edition. This GA notice is a first-time source; it was updated 10/30 and is not in the archive.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-01; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Large unstructured file estates with unknown stale, duplicate and over-shared data, and pressure to make data AI-ready under records rules.
- Stakeholders
- CIO or IT director
- Records manager or university archivist
- Information security officer
- Storage and server team lead
- Data governance or privacy office
- Discovery questions
- Which file shares and databases hold the most unmanaged data today?
- What retention schedules and litigation holds apply to those shares?
- Is Kubernetes in house, or would a single VM be the only option?
- Which AI clients, if any, are approved to touch institutional metadata?
- Is Microsoft Entra ID the identity provider for staff who would query collections?
- Value hypothesis
- A scoped scan can show where cold, stale, duplicate and open-permission data sit, giving retention and security teams a fact base. Savings depend on what owners agree to archive or delete.
- Potential engagement
- Fixed-scope AIDE pilot: single-node deployment, scan of a defined share set, findings review with data owners, and a retention-policy mapping workshop.
- Unsupported claims
- Guaranteed storage or cost savings
- Semantic search, vectorization or RAG features in 1.3
- Production scale or accuracy at SLED agencies
- Any SLED customer reference
Pre-sales engineering
Role takeaway
- Fit
- Fits estates with NFS or SMB shares, supported database servers or StorageGRID where in-place metadata cataloging is wanted.
- Architecture and integration
- Single-node on a customer Linux VM with installer-managed k3s, or multi-node on customer RKE2, OpenShift or AKS; managed through NetApp Console and connected to the Console agent.
- Prerequisites
- Linux VM or Kubernetes cluster sized to file count
- NetApp Console account and Console agent
- Network access to data sources over NFS or SMB
- Microsoft Entra ID if governed MCP access is used
- Constraints
- Single-node has no high availability
- Multi-node cluster is customer-provisioned and operated
- Sizing above about 6 billion files needs custom sizing with NetApp
- Online installation depends on cloud endpoints (see requirements resource)
- Security
- Catalog holds paths, owners and permissions; restrict viewer and admin roles, scope MCP collections by Entra ID, and log AI-client queries and tag actions.
- Proof of value
- Scan a representative share and compare file counts with storage-system reports
- Have data owners sample-check cold, stale and duplicate findings
- Test that an unscoped Entra ID user cannot search a restricted collection
- Measure actual scan rate against the documented range for the chosen size
Delivery
Role takeaway
- Work
- Deploy single-node AIDE, connect agreed sources, run scans, review findings with owners and configure monitoring policies without destructive actions.
- Dependencies
- Console agent deployed and reachable
- Records retention schedule
- Data owner availability for review
- Change approval for VM or cluster
- Ownership
- Infrastructure team operates the VM or cluster; records management owns retention policies; security owns MCP access scopes.
- Skills and adoption
- Storage admins need Console and AIDE policy skills; multi-node adds Kubernetes operations; records staff need training on interpreting findings.
- Governance checkpoints
- Records officer approval before any delete or archive policy
- Security review of MCP client list and Entra ID scopes
- Privacy review of metadata exposure for sensitive shares
- Acceptance criteria
- Proposed: scan completes for all agreed sources within the pilot window
- Proposed: data owners confirm a sample of findings as accurate
- Proposed: no destructive action executed without recorded approval
- Risks
- Automated deletion conflicting with public-records or legal-hold duties
- Feature expectations based on the 2025 announcement rather than 1.3
- Unplanned Kubernetes operating burden for multi-node
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Two deployment modes. Single-node runs on a customer-supplied Linux VM, with k3s installed by the installer and no high availability. Multi-node runs on a customer-operated RKE2, OpenShift or AKS cluster with three to nine nodes. Console guides installation but does not provision infrastructure. AIDE reads NFS and SMB sources and StorageGRID, so data stays in place.
Governance
Who approves, reviews and stays accountable for outcomes?
Delete and archive actions plus saved policies mean AIDE can enforce retention rules. Public-records and litigation-hold obligations must be mapped before any automated deletion. MCP access is scoped by Entra ID per collection, so AI-client access becomes a governed decision rather than a default.
Security and privacy
What data, permissions and controls need testing?
Catalog metadata includes file names, owners, permissions and paths, which can be sensitive in HHS, public safety or student-records contexts. Open-permission findings are useful for least-privilege reviews. Connecting third-party AI clients through MCP extends who can query metadata and tag files.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Adds a data-hygiene operations role for storage and records staff; multi-node needs Kubernetes skills many SLED IT teams lack. No accessibility findings in the source.
Procurement
What should contracts, pricing and exit terms secure?
GA allows a standard purchase and support conversation; pricing and licensing are not in the source and need to be quoted. Cooperative contracts used for NetApp hardware may not cover AIDE without confirmation.
Operating model
Which teams own the service once it runs?
Moves storage teams toward continuous data-lifecycle governance with records officers, rather than one-off cleanup projects.
Publication history
- 2026-10-31NetApp · Issue 033 resources
Stable resource ID: netapp-aide-1-3-first-ga-release-whats-new-20261030