QORM vs the Agentic-UI Ecosystem
Agentic UI and generative UI in 2026 split into two layers: protocols that describe how an agent talks about UI, and complete platforms that render, sign, deliver, and package real apps. QORM is the second kind.
Two Layers, Not Two Competitors
The agentic-UI landscape has matured into a stack. At the protocol layer, standards define how an AI agent describes and transports UI: Google's A2UI specifies an agent-to-client UI message format, CopilotKit's AG-UI defines the agent-to-frontend event stream, MCP Apps let tools return interactive UI resources, and the Vercel AI SDK popularized generative UI inside React apps. These projects answer the question: how does an agent say "render this"?
At the platform layer, something has to actually render that UI, keep it in sync between the human and the agent, sign it for integrity, deliver updates over the air, and package it for web, iOS, Android, desktop, and mini-programs. That is what QORM is: a pure-Go runtime where the app itself is language-neutral JSON, exposed live to AI agents over 20+ MCP tools while humans watch and interact in the same session.
So "A2UI vs AG-UI vs QORM" is usually the wrong framing — protocols and platforms solve different problems and can be complementary. The detail pages below map exactly where each one stops and where QORM starts.
QORM vs Google A2UI
A2UI is an LLM-friendly agent-to-client UI message format — a protocol, deliberately silent on rendering, delivery, signing, and packaging. QORM is the full runtime and delivery platform underneath.
Read the comparisonQORM vs AG-UI (CopilotKit)
AG-UI is an agent-to-frontend event and transport protocol; CopilotKit is a React framework around it. In QORM the runtime is the app — agents mutate the live UI directly over MCP while humans watch the same session.
Read the comparisonThe Ecosystem at a Glance
| Project | Layer | What it does | What it leaves to you |
|---|---|---|---|
| Google A2UI | UI message protocol | Defines how an agent streams component trees to a client | Renderer, delivery, signing, packaging |
| CopilotKit AG-UI | Event / transport protocol | Streams agent state, messages, and tool calls into a frontend | The app runtime, persistence, distribution |
| MCP Apps | Tool-UI resource spec | Lets MCP tools return interactive UI to a host | Cross-platform packaging, OTA, app lifecycle |
| Vercel AI SDK | Generative UI toolkit | Renders model output as React components in web apps | Non-React targets, signing, native delivery |
| QORM | Complete platform | Renders JSON apps with a dual renderer — an HTML/DOM renderer for web/WebView and a native canvas software renderer (pure Go, pixel-based, also compiles to WASM) for desktop and browser games — signs bundles, delivers OTA, packages for web/iOS/Android/desktop/mini-program, exposes 20+ MCP tools to agents | Nothing — it is the runtime |
If you already have a renderer and a delivery pipeline, a protocol may be all you need. If you need a complete, signed, OTA-updatable cross-platform app that agents can drive live alongside humans, that is the layer QORM occupies.
See the Platform in Action
QORM is open source under the MIT license. Run an example app locally, then drive it from your favorite agent over MCP.
github.com/qorm/qorm