LearnNewsExamplesConnect

The Application Engine for the AI Era

Traditional frameworks compile away. Engines stay alive.

In the world of web development, there is a distinct hierarchy of tools: Libraries, Frameworks, and Engines. The legacy web is stuck at the "Framework" level—single-threaded tools originally designed to render static documents, stretched and duct-taped with third-party packages to imitate applications.

Neo.mjs is not a framework. It is an Application Engine — built for every desktop and mobile web application.

Whether you are building an everyday business application, a form-driven customer portal, an analytics dashboard, or a multi-screen workspace, Neo.mjs gives you a complete, batteries-included platform out of the box (461 classes across 39 namespaces: 100,000-row buffered grids, a 26-field form engine with nested forms, calendars, dashboard docking, tabs, dialogs, toolbars, menus, lists, trees, 8 layouts, data stores, state providers, routing, and 38 main-thread addons). It builds everything teams build with Angular, React, or Vue today — at feature parity with Ext JS and generations ahead of it in architecture — while its multi-threaded runtime and live possession interface effortlessly power use cases where legacy single-threaded frameworks fatally bottleneck: 40k+ ops/sec real-time data streams, multi-window applications sharing one SharedWorker heap, and complex workspaces that never freeze.

The Paradigm Shift: Why the Industry Must Adapt

To understand why Neo.mjs is the only platform truly ready for the AI era, we must examine the "original sin" of the legacy web: transient DOM rendering locked to a single Main Thread.

1. The Persistent Scene Graph

Legacy Frameworks are Transient Blueprints. When a traditional framework renders a component, the output (the DOM) is just a visual snapshot. The "component" itself effectively ceases to exist. It is melted down. If state changes, the framework must aggressively diff and re-melt the entire blueprint against the bottlenecked Main Thread.

Neo.mjs is a Persistent Scene Graph. Components in Neo.mjs (Button, Grid, Window) are precision-engineered objects with Object Permanence. They exist as live, stateful nodes within a dedicated App Worker. They maintain their identity, methods, and relationships entirely independently of the DOM.

2. The JSON-First AI Synergy

Large Language Models (LLMs) natively excel at reading, reasoning over, and generating JSON. Since its initial GA in November 2019 — years before trends like Vercel's json-render — Neo.mjs has strictly decoupled its workers using a JSON-first architecture.

Because workers physically cannot touch the live DOM, the entire application — component trees, layouts, data stores, state providers, and VDOM deltas — is represented as pure, serializable JSON blueprints (static config and toJSON()). Because LLMs excel at JSON, and Neo.mjs has been JSON-first since 2019, LLMs excel at building with Neo.mjs — especially when paired with Neo's AI infrastructure. An AI model doesn't have to wrestle with HTML quirks, JSX compilation steps, or useEffect dependency arrays; JSON is its native tongue.

3. Simpler for Humans, Native for Coding Agents

Multi-threading in Neo.mjs does not add complexity to your application code — it removes it. The engine handles worker orchestration, cross-window state synchronization, and VDOM diffing behind the scenes.

In the AI era, LLMs and Neo.mjs are a natural match.

Because LLMs excel at JSON and Neo.mjs has been JSON-first since its November 2019 GA, coding agents write Neo.mjs components, layouts, forms, grids, and state providers in the same declarative shape a human does, and check the result in the running app through the Neural Link — especially when paired with our AI infrastructure. And instead of the industry's Rung-1 Command-and-Control model — a human "orchestrator" babysitting stateless sub-agents that forget every session, which What Is Neo.mjs? shows is architecturally capped at rung one — Neo.mjs asks the inverted question: how can software become a body an AI inhabits, and a team of AIs can be trusted to run?

Built around Model Experience (MX), Object Permanence, and JSON-first worker contracts, Neo.mjs lets your coding agent — or a standing team of accountable, named AI peer maintainers across rival labs (Claude, Gemini, GPT) with persistent institutional memory, cross-family review rights, and a human gardener at the merge gate — build and evolve applications that are physically impossible in legacy frameworks.

Runtime Embodiment: The Map IS the Territory

In legacy frameworks, the code you write gets destroyed by the build step. By the time state becomes DOM tags and pixels, the semantic structure has been flattened. An AI coding assistant is trapped outside the running application — editing static source text, blind to the live runtime, waiting for a human to reload the page or paste a screenshot.

In Neo.mjs, the map IS the territory — and the AI inhabits it.

Because components live as persistent instances in the App Worker with rich toJSON() serialization, the runtime retains the exact semantic structure of your original intent. Through the Neural Link (Possession Interface), your coding agent (Claude Code, Codex, Gemini) — or an AI maintainer team — reaches directly into your living, running application across multiple browser windows:

  • Live Scene Graph & State Introspection: Query "Get me the Grid with reference 'sales-report'" and the Engine replies with the live instance — its 10,000 store records, its hierarchical state.Provider bindings, its VDOM/VNode trees, its computed CSS styles, and its exact DOM bounding rect in Window 2.
  • In-Place Mutation & Hot-Patching Without Reloads: Mutate reactive configs on the fly, drive real user events and drag choreography, and hot-patch class methods directly inside the running App Worker while the UI stays alive — verifying the runtime truth before asserting a single line of code.

Joined through the Neural Link to the Brain (neomjs/neo-agent-brain — the Agent OS housing the Native Edge Graph, Memory Core, Dream Pipeline, Knowledge Base, and GitHub Workflow) and the Fleet Manager (neomjs/neo-agent-institution), the Body and Brain form one self-evolving organism capable of diagnosing live bugs, inventing new design patterns, and evolving the Engine itself in public.

Conclusion: Build the Impossible

Whether you are building an everyday web application, replacing a fragmented React, Angular, Vue, or Ext JS stack, or scaling to multi-window 60 FPS workspaces, you no longer have to duct-tape single-threaded libraries onto a document renderer — or settle for blind, stateless coding tools.

Neo.mjs gives you a complete, zero-builds Application Engine that runs off the main thread, a Possession Interface where AI peers see, touch, and hot-patch your running application live across windows, and the institutional substrate where human gardeners and accountable AI maintainers build together.