CopilotKit adds AG-UI and A2UI support to connect Open Agent Spec agents to generative UI
CopilotKit added AG-UI and A2UI support so agents defined with open specs can stream interactive UI directly to the frontend. Use it to reduce custom glue between backend agent runtimes and user-facing interfaces across stacks.

TL;DR
- CopilotKit says it has added support for AG-UI and A2UI, so agents defined with Open Agent Spec can stream interactive, generative UI to the frontend instead of relying on custom transport code.
- The core pitch in the launch thread is portability: define the agent once, emit A2UI messages, and let an AG-UI endpoint feed live events into CopilotKit across different runtimes and UI stacks.
- CopilotKit also says in its product demo that the same stack now covers AG-UI, A2UI, and MCP Apps, positioning it as a bridge between backend agents and interactive app surfaces.
- A follow-up demo in the starter-template post ties the launch to a concrete workflow: scaffold with
npx copilotkit@latest create, connect to a backend agent, and render interactive charts or other generated UI in-app.
What shipped
CopilotKit's announcement centers on a three-part integration: Open Agent Spec for agent definition, A2UI for generative UI messages, and AG-UI for streaming those events to the frontend. In the launch thread, the company describes the result as "no custom transport or glue code" and says the goal is "plug-and-play portability" across runtimes, UI frameworks, and observability tooling.
A second post makes the developer-facing scope clearer. According to the demo post, teams can bring "interactive, generative UI" into their own apps and copilots with support for AG-UI, A2UI, and MCP Apps, which suggests CopilotKit is treating these protocols as interchangeable integration surfaces rather than one-off adapters.
How the integration works in practice
CopilotKit's thread lays out the runtime path step by step: a backend agent defined with Agent Spec emits tracing events such as tool calls, state updates, and progress; those events are exposed through an AG-UI-compatible endpoint; CopilotKit subscribes to that live stream; and A2UI messages are rendered as interactive UI, with user actions sent back into the agent loop through AG-UI integration steps.
That makes this more than a UI demo. The starter-template post says CopilotKit "connects directly to any agent in the backend" and pairs the launch with a scaffold command, while the attached demo shows an interactive chart rendered from the generated interface. Together, the posts frame AG-UI and A2UI support as an attempt to standardize the path from agent state and tool execution to user-facing components, instead of rebuilding that bridge for each app.