AGenUI - A native A2UI framework open-sourced by Gaode Maps and Alibaba. - AiBoss

What is AgenUI?

AGenUI is the industry's first open-source, edge-cloud integrated native A2UI framework, jointly launched by Amap and Alibaba's Qianwen C-end application team, covering iOS, Android, and HarmonyOS. Built on the Google A2UI open protocol, the framework enhances edge-side native rendering capabilities, allowing the output of the AI Agent to be directly converted into interactive native cards. This eliminates the need to write separate UI code for different platforms, propelling AI applications from text-based interaction to generative UI interaction.

AGenUI's main functions

Native rendering across three platformsA single codebase can run on iOS, Android, and HarmonyOS simultaneously, rendering native components directly on the device side.
A2UI Protocol Full ImplementationBased on the Google A2UI protocol, it defines a standard way to describe the interface of the model and adds the execution capability on the client side.
Enrich the component ecosystemIt includes 22 basic components and 45 CSS properties, and supports complex components such as Chart, RichText, Lottie, PageViewer, and audio/video players.
Streaming Real-Time RenderingIt adopts a Streaming-first architecture, where components are generated and mounted as they are rendered, combined with differential updates and asynchronous rendering.
Theme Design SystemAutomatic mapping from semantic descriptions to brand styles is achieved through Design Token, supporting both dark and light modes.
3D customization of componentsIt supports flexible customization and overriding at the component level, function call level, and theme level.

Technical Principles of AGenUI

End-to-cloud integrated architectureOn the cloud side, Agent Skill generates JSON data that conforms to the A2UI specification, reducing the token consumption and output uncertainty of large models; on the client side, cross-platform C++ Core is used to uniformly handle protocol parsing, state management and layout calculation, and renders it directly as native components on all three platforms, ensuring a consistent experience across multiple platforms from the bottom up.
Streaming-first architectureIt supports mounting components as soon as they arrive, enabling "generation and rendering simultaneously"; combined with a mechanism for minimizing node differential updates and asynchronous rendering in independent threads, high-frequency incremental updates will not block the main thread, ensuring a smooth 120fps animation experience.
Native cross-platform kernelUsing C++ Core as a cross-platform foundation, it directly calls the native rendering pipeline on various platforms, without going through WebView or a self-drawing engine, to achieve ultimate first-frame rendering performance and native interactive experience.

AGenUI's core advantages

One codebase for three platformsA single codebase covers iOS, Android, and HarmonyOS, completely eliminating the cost of redundant cross-platform UI development.
Native rendering on the deviceIt directly calls the native rendering pipeline of each platform, which greatly reduces the first frame time and achieves a smooth animation of 120fps, providing an experience far superior to the WebView solution.
Streaming Real-Time PresentationIt adopts a Streaming-first architecture, where components are generated and mounted as they are rendered, allowing users to start interacting without waiting for a complete response.
Differential Asynchronous UpdateMinimize node differential updates and combine them with asynchronous rendering on independent threads. High-frequency incremental updates do not block the main thread, ensuring smooth performance even in complex scenarios.
Built-in design systemBy using Design Token, the semantics are automatically mapped to the brand. The UI generated by the Agent can be directly aligned with the product's visual standards and supports dark/light modes.
Minimalist access experienceUnified API design allows SDK integration to be completed with just a few lines of code, significantly reducing the barrier to Agent UI development.

AGenUI project address

Project official websitehttps://genui.amap.com/
GitHub repositoryhttps://github.com/AGenUI/AGenUI

Comparison of similar products with AGenUI

Comparison Dimensions	AGenUI	AG-UI (CopilotKit)	React Native
Product Positioning	A cloud-edge integrated native UI generation framework for AI agents, enabling large model outputs to be directly transformed into interactive native interfaces.	The front-end Agent-UI real-time interaction protocol enables bidirectional data flow between LLM and front-end components, focusing on web scenarios.	Traditional cross-platform mobile application development frameworks use JavaScript to write logic and native components to render the UI, which is not designed for Agent scenarios.
Protocol basis	Based on the Google A2UI open protocol, a standard JSON format for the model description interface is defined, and native execution capabilities on the client side are added.	The self-developed AG-UI protocol drives the state updates of front-end components through event flow, realizing two-way binding between LLM and UI.	Without a specific Agent protocol, developers need to handle the parsing and mapping logic of LLM output to the UI themselves.
Supported Platforms	It covers iOS, Android, and HarmonyOS platforms, allowing a single codebase to run across all platforms.	It primarily uses the Web ecosystem and relies on front-end frameworks (React/Vue/Angular). Mobile applications require WebView or PWA.	It supports iOS and Android, and can be extended to Web and desktop through community solutions, but it does not support native HarmonyOS.
Rendering method	Native rendering on the client side, cross-platform C++ Core directly calls the native rendering pipeline of each platform, without WebView intermediary.	The front-end framework is independent of rendering; it ultimately relies on the rendering engine of the browser or WebView, and is not directly output from native components.	Native rendering, with JavaScript bridging to call platform native components, results in performance close to that of a pure native application.
AI Native Design	Designed from scratch specifically for Agent scenarios, components, layouts, and interactions are all optimized around the generative output of large models.	Designed specifically for Agent scenarios, it provides front-end-level LLM state management and component linkage mechanisms.	A general-purpose application development framework with no built-in Agent or LLM interaction capabilities; all must be completely encapsulated by the developer.
Streaming generation	It adopts a Streaming-first architecture, where components are generated and mounted as they are rendered, combined with differential updates and asynchronous rendering.	It supports event-driven programming, and the LLM output can be mapped to the front-end state in real time, but component rendering is limited by the performance of the front-end framework.	Native fluid UI generation is not supported. To achieve a similar effect, developers need to build a complex state management solution themselves.
Component ecosystem	It includes 22 basic components and 45 CSS properties, covering complex components such as Chart, RichText, Lottie, and audio/video players.	It relies on existing component libraries in the front-end ecosystem and binds LLM capabilities through the protocol layer. The richness of components depends on the selected front-end framework.	It has a large community component ecosystem, but the components need to be manually orchestrated and cannot be dynamically generated and combined by the Agent.
Performance	The first frame rendering time has been significantly reduced, supporting smooth 120fps animation, and high-frequency incremental updates do not block the main thread.	Performance is limited by the rendering efficiency of the browser or WebView, and frame drops are likely to occur in complex animations and high-frequency update scenarios.	Native rendering offers excellent performance, but JavaScript bridging communication incurs overhead, requiring additional optimization for complex scenarios.
Access method	The unified API can be integrated with the SDK with just a few lines of code, and the A2UI data generated on the cloud side will be automatically mapped to the native interface on the client side.	The SDK needs to be included in the front-end project and the AG-UI protocol adaptation layer needs to be configured, which requires certain front-end technology stack.	A complete development environment needs to be set up, and code for bridging JSX components and native modules needs to be written, resulting in high integration costs.
Open source license	open source	MIT Open Source	MIT Open Source

Application scenarios of AgenUI

Smart Assistant Interactive CardsIn complex scenarios such as map navigation, local services, and travel services, the Agent generates interactive native cards in real time, allowing users to complete operations such as querying, booking, and payment without having to navigate to a different page.
Dynamic data visualizationThe Agent generates native chart components such as line charts, bar charts, and pie charts in real time based on the dialogue context, transforming abstract data into an intuitive and visual native interface on the client side.
Rich Media Content DisplayIt supports the dynamic generation of rich interactive content such as mixed text and images, carousels, Lottie animations, and audio and video playback, meeting the multimedia presentation needs of marketing, information, education and other scenarios.
Enterprise-grade cross-platform agent applicationsFor B-end intelligent agent products that need to be deployed on iOS, Android, and HarmonyOS simultaneously and have high requirements for performance, visual consistency, and smooth interaction, this significantly reduces the cost of multi-platform adaptation.