Security

Secure by design

Neutron AI gives teams scoped memory, tenant boundaries, deletion-aware recall, and controlled integration surfaces for agent platforms.

Tenant and scope boundaries
Every memory operation is tied to explicit tenant and scope identifiers, so user, workspace, project, session, and agent context can stay separated by design.
Access control
API keys, admin sessions, and integration permissions are designed around least-privilege access. Sensitive management actions are kept behind authenticated control surfaces.
Deletion-aware memory
Forget flows are first-class. Deleted memories are marked so recall, cache refresh, compaction, and archive workflows do not reintroduce removed context.
Enterprise controls

Control how memory enters, moves, and leaves your product.

Scoped memory
Data handling

Memory is stored as scoped product context, not open-ended chat history.

Neutron AI is built for bounded remember, recall, reflect, context, and forget operations. Applications decide what to send, which scope it belongs to, and when it should be deleted.

Provider boundary
Model boundaries

Provider credentials stay with your application.

Neutron AI provides memory APIs, SDKs, CLI workflows, and MCP tools that sit beside your model provider. Your app remains responsible for model selection, provider credentials, and prompt assembly policy.

No training use
Training policy

Customer memory is not positioned as model-training data.

The public product model is memory infrastructure for retrieval and context packaging. Customer memory should be treated as application data, not as training material for general-purpose models.

Data storage

Choose the storage profile that matches your product and review requirements.

Neutron AI supports global, regional, and jurisdictional placement profiles so teams can plan memory storage around availability, latency, and regulated deployment needs.

Global storage

Default placement for globally distributed apps. Memory is scoped to your tenant and product boundaries while the platform optimizes for broad availability and low-friction integration.

Best for public apps, prototypes, and general agent products.

Regional storage

Regional placement preferences can be used when teams want memory and archive workflows optimized around broad regions such as North America, Europe, or Asia-Pacific.

Best-effort regional placement for latency and operational preference, not a standalone legal residency commitment.

Jurisdictional storage

Jurisdictional profiles are intended for regulated Enterprise deployments where data storage needs to align with supported jurisdictional boundaries and stricter review paths.

Supported profiles include EU and FedRAMP-oriented deployment review, subject to contract and configuration.

Placement is part of deployment design
Regional placement preferences do not replace legal, contractual, or account-level controls. Enterprise teams can review jurisdictional requirements, retention, access, and support obligations before regulated production use.

Monitoring and abuse resistance

Public memory infrastructure needs predictable contracts, bounded usage, and auditable controls. Neutron AI is designed so teams can observe memory behavior without exposing sensitive payloads.

  • Usage analytics for memory operations and retained memory
  • Audit-friendly records for sensitive administrative actions
  • Rate limits and bounded API contracts for public integration surfaces
  • Structured error responses with stable codes for safer automation

Compliance-ready packaging

Security review should be practical for teams adopting agent memory. Pro and enterprise plans package the controls procurement teams expect for real workloads.

  • Data protection review support for enterprise procurement
  • Configurable retention and deletion workflows for production plans
  • Global, regional, and jurisdictional storage profiles for production deployment planning
  • SSO/SAML, audit exports, and dedicated review paths for enterprise customers
  • Public documentation for REST, SDK, CLI, and MCP integration behavior
FAQ

Security questions, answered directly.

These answers are written for external teams evaluating Neutron AI for SDK, API, CLI, and MCP integration.

Ready to build?

Bring secure memory into your agent platform.

Start with scoped recall, deletion-aware workflows, and provider-safe context packaging before expanding into production governance.