Memory for every agent.
Give every agent fast, permissioned memory.
Memory infrastructure for agentic products.
Neutron AI gives teams a clear way to add durable memory to agents while keeping scope, privacy, deletion, and provider boundaries explicit.
Scoped memory
Give agents memory without losing boundaries.
Store memories by tenant, user, project, team, agent, or session so every recall path has an explicit permission and consistency boundary.
Context packs
Turn recall into reusable model context.
Build bounded context packs that combine relevant memories, summaries, scores, cache policy, and safe metadata for provider calls.
Deletion safety
Keep memory auditable and deletion-aware.
Use tombstones, scope versions, cache freshness checks, and audit events so deleted memory does not silently return through stale context.
From first memory to production recall.
The core workflow stays simple: remember scoped context, recall what matters, pack it for a model call, and keep the memory lifecycle auditable.
- 1
Define tenant, scope, privacy class, and agent policy.
- 2
Remember product, user, project, tool, and workflow context.
- 3
Recall the right memories and build an AgentContextPack.
- 4
Reflect, compact, forget, and audit memory over time.
Built for agent teams, not one-off chats.
Start with a single assistant, then reuse the same memory model across products, workflows, tools, and agent runtimes.
Connect memory to the stack you already use.
Neutron AI is designed to sit beside your model provider and app runtime. Provider credentials stay with your application.
REST API
https://neutronai.dev/api
TypeScript SDK
@neutron-ai/sdk
Python SDK
neutron-ai
PHP SDK
neutron-ai/sdk
Go SDK
github.com/neutron-ai/sdk-go
Rust SDK
neutron-ai
CLI
@neutron-ai/cli
MCP tools
https://mcp.neutronai.dev
Memory with explicit controls.
Scope validation, deletion behavior, redaction, audit events, and safe logging are part of the product model. Public docs focus on purpose, values, and integration boundaries without exposing private platform details.
Pricing built for agent memory.
Plans are based on memory operations and retained memory, with SDKs, CLI, MCP, admin seats, and context-cache access bundled in.
- 50k memory operations per month
- 256 MB retained memory
- SDKs, CLI, REST, and MCP
- Basic usage analytics
- Community support
- 500k memory operations per month
- 2 GB retained memory
- 3 projects
- Context cache included
- $35 per extra 1M memory operations
- 4M memory operations per month
- 20 GB retained memory
- 15 projects
- Admin seats bundled
- Audit exports and custom retention controls
- Custom included operations and memory
- Regions, SLAs, SSO/SAML, and compliance contracts
- Dedicated support
- Negotiated overages
- Custom annual contracts and Stripe prices