# CopilotKit Expands Agent UI Framework Across React, Vue, Angular > CopilotKit has released a multi-platform frontend framework for building agent-native applications, supporting React, Angular, Vue, and React Native. The SDK provides generative UI, shared state, and human-in-the-loop workflows, and the company also maintains the AG-UI Protocol, which has been adopted by Google, LangChain, AWS, and Microsoft. - Canonical URL: https://agentry.press/news/copilotkit-expands-agent-ui-framework-across-react-vue-angular/ - Type: News - Published: 2026-06-15 - By: agentry - Tags: copilotkit, agent-ui, ag-ui-protocol, generative-ui, human-in-the-loop, frontend-agents --- ## What CopilotKit Is CopilotKit began as a React library and has since expanded into a multi-platform agentic framework spanning web, mobile, and messaging surfaces [1]. The SDK targets frontend engineers building agent-native applications, providing tooling for React, Angular, Vue, and React Native. The company also maintains the AG-UI Protocol, a wire protocol that separates agent logic from UI rendering and has been adopted by Google, LangChain, AWS, Microsoft, Mastra, PydanticAI, and others [1]. The framework's stated design goal is that a single agent backend can power a web application, a mobile application, and a team's Slack workspace without changes to the underlying agent logic [1]. ## Core Features CopilotKit exposes six primary capabilities to application developers [1]. **Chat UI** provides a customizable chat interface that supports message streaming, tool calls, and agent responses. **Backend Tool Rendering** enables agents to invoke backend tools that return UI components rendered directly in the client, keeping rendering logic server-side while the result surfaces in the browser or app. **Generative UI** allows agents to generate and update UI components dynamically at runtime, driven by user intent and agent state rather than static templates. **Shared State** introduces a synchronized state layer that both agents and UI components can read from and write to in real time, giving the agent and the interface a common view of application data. **Human-in-the-Loop** lets agents pause execution to request user input, confirmation, or edits before continuing a workflow. This capability extends into messaging platforms: the framework supports human-in-the-loop approvals directly inside Slack threads and Microsoft Teams channels [1]. **Self-Learning**, currently in early access, enables agents to improve continuously from user feedback through in-context reinforcement learning, which CopilotKit refers to as CLHF [1]. CopilotKit is onboarding teams for early access on a selective basis [1]. ## The AG-UI Protocol The AG-UI Protocol functions as the wire layer between an agent backend and any frontend surface [1]. By handling the protocol translation at this layer, agent logic remains unchanged regardless of which framework or platform renders the output. The protocol has attracted adoption from a broad set of organizations, including Google, LangChain, AWS, Microsoft, Mastra, and PydanticAI [1]. The separation of concerns the protocol enforces means that CopilotKit can implement the UI-layer rendering for each supported framework independently, while the agent communicates through a single standardized interface. ## Architecture and Integration The architecture follows a two-layer model. AG-UI handles the wire protocol between the agent and the client, and CopilotKit handles the UI rendering layer for each specific framework [1]. A team deploying an agent to both a React web application and a React Native mobile application uses the same agent backend in both cases. The CopilotKit SDK adapts the output to the target surface. Beyond browser and mobile targets, the framework supports deployment to Slack as first-class Slack applications, with threads, tool calls, and human-in-the-loop approvals available inside channels. Microsoft Teams integration brings the same agentic workflow capabilities to enterprise messaging environments [1]. Both messaging integrations are listed as early access, with teams being onboarded on a rolling basis. ## Who the Framework Targets CopilotKit positions the SDK for frontend engineers building agentic products and for teams that need to deploy the same agent logic across multiple surfaces without maintaining separate agent implementations per platform [1]. The multi-platform scope addresses a common operational problem: agent logic duplicated or forked to serve web, mobile, and messaging audiences separately. ## FAQ **Q. Which frontend frameworks does CopilotKit currently support?** The SDK supports React, Angular, Vue, and React Native [1]. The company describes the framework as extending "beyond the browser" through Slack and Microsoft Teams integrations. **Q. Is the Self-Learning feature available to all users?** No. Self-Learning, which uses in-context reinforcement learning (CLHF), is listed as early access, and CopilotKit is onboarding teams on a selective basis [1]. **Q. Does adopting CopilotKit require changing existing agent logic?** According to the project documentation, agent logic remains the same across surfaces. AG-UI handles the wire protocol and CopilotKit handles the UI layer for each framework, so the agent backend does not need to be modified per platform [1]. **Q. Which organizations have adopted the AG-UI Protocol?** The protocol has been adopted by Google, LangChain, AWS, Microsoft, Mastra, and PydanticAI, among others [1]. **Q. Are the Slack and Microsoft Teams integrations generally available?** Both messaging integrations are in early access. CopilotKit states it is currently onboarding teams for these surfaces [1]. ## Key Takeaways - CopilotKit has expanded from a React library into a multi-platform framework covering React, Angular, Vue, React Native, Slack, and Microsoft Teams [1]. - The AG-UI Protocol separates agent logic from UI rendering and has been adopted by Google, LangChain, AWS, Microsoft, Mastra, and PydanticAI [1]. - Six core capabilities include Chat UI, Backend Tool Rendering, Generative UI, Shared State, Human-in-the-Loop, and Self-Learning [1]. - A two-layer architecture (AG-UI for the wire protocol, CopilotKit for per-framework UI rendering) allows one agent backend to serve multiple frontend surfaces without modification [1]. - The Self-Learning feature, using in-context reinforcement learning (CLHF), is in early access and requires team onboarding [1]. ## Sources 1. https://github.com/CopilotKit/CopilotKit