AWS is the most comprehensive cloud platform — but leveraging its full power as a developer remains hard. CDK codifies infrastructure, yet IAM policies, service integrations, and operational complexity still create hurdles. AWS Amplify Gen 2 bridges this gap with a developer-first, code-first approach that makes building full-stack applications fast. But Amplify today covers only a slice of what real applications need.
Every application sends emails. Every application needs teams collaborating on shared resources with proper permissions. Every application needs a notification system. Every application needs to work when the user goes offline. And increasingly, every application needs AI agents that understand your data.
Amplify Overtone closes this gap. It extends Amplify Gen 2 with the capabilities that production apps demand — declared in your schema, generated as CDK constructs, and consumed through type-safe client APIs. We're shipping five new categories: Email, Collaboration, Notifications, Local-First, and Agent.
| Package | npm | Description |
|---|---|---|
@nxsflow/amplify-overtone |
 |
Backend: schema builder, CDK constructs, resolvers |
@nxsflow/amplify-overtone-client |
 |
Frontend: sync engine, IndexedDB storage, headless APIs |
Transactional email via Amazon SES — custom domains, named senders, built-in templates, and DNS automation.
- #4
defineEmail()— SES email infrastructure — in progress - #5
n.email()— schema-defined email actions — declarative email sending from the schema with typed arguments, sender selection, and AppSync resolver generation - #6 Client API for email actions — type-safe
client.actions.*for invoking email actions from the frontend
CRDT-based conflict-free editing powered by Automerge — dual-table storage, real-time operation broadcasting, authorization inheritance, and a collaborative session API.
- #7 Collaborative data —
.collaborative()models with OpsLog, compaction, version vectors, presence, and first-class authorization fields (n.owner(),n.editors(), etc.)
Offline-capable IndexedDB storage via cr-sqlite — zero-latency reads, background sync, offline mutation queues, and lease-based authorization with role-based TTLs.
- #8 Local-first storage —
.localFirst()models with sync engine (Web Worker), offline queue, and lease expiry lifecycle
Full notification subsystem — in-app notifications via AppSync subscriptions, OS-level push via APNs/FCM/Web Push, email channel bridging, and user preference management. Headless client API for toasts, badge counts, and notification settings.
- #3 Notification subsystem —
defineNotifications()factory,n.notify()schema actions, Step Function orchestrator, and headlessuseNotifications()/useNotificationPreferences()hooks
AI agents powered by Amazon Bedrock AgentCore — defineAgent() provisions the runtime (Strands SDK), memory, and embeddings pipeline. n.agent() declares agent actions in the schema with model references that auto-generate typed tools. Your data model IS your agent's capability surface.
- #9
defineAgent()+n.agent()— AI agent infrastructure and schema actions — AgentCore Runtime, auto-generated tools froma.ref()model grants, resource-scoped memory and embeddings, AgUI protocol for streaming and conversational interaction
pnpm add @nxsflow/amplify-overtoneThe packages are not yet published to npm. Check back soon.
We welcome contributions! Each roadmap issue above is a good starting point — they contain detailed design specs, API examples, and architecture decisions. See CONTRIBUTING.md for development setup, testing, and release workflow.