Bring your Agent Spec JSON to AG-UI frontends for real-time, stateful, and tool‑aware experiences.

We’re excited to announce that Open Agent Specification (Agent Spec) and AG-UI (from CopilotKit) now work together to connect portable backend agent orchestrations with interactive, frontend experiences.

Define agents declaratively once, run them on compatible runtimes including LangGraph, WayFlow (and others coming soon), and connect to AG-UI frontends without custom integration glue.

This reduces dependence on any single framework, accelerates UI delivery, and creates a clean handoff between agent orchestration and the application experience while Agent Spec Tracing lets standardized events flow to frontends and downstream observability and evaluation tools via pluggable hooks.

Diagram for how Agent Spec fits in the AI Agents ecosystem with AG-UI: Agent Spec configurations are run by agentic frameworks, and Agent Spec tracing emits standardized events listened by the AG-UI protocol to interact with the frontend

Agent Spec and AG-UI in brief

  • Agent Spec: Open Agent Specification (Agent Spec) is a framework-agnostic, declarative language for describing AI agents, structured workflows, and multi-agent patterns. It captures behavior, tool use, and orchestration logic in a portable configuration, so developers can define once and execute across compatible runtimes without rewriting logic.
  • AG-UI: AG-UI (Agent–User Interaction Protocol) is an open, lightweight, event-based protocol that standardizes how agents connect to user-facing applications. It provides a common event stream for messages, tool activity, and UI state so teams can build reliable, debuggable agentic experiences without bespoke wiring. Frontends can be built with any AG-UI–compatible client (for example, CopilotKit’s React components).

What this integration enables

This integration connects portable agent definitions to modern, event-driven frontends, so you can move from configuration to interactive experience quickly.

  • Seamless deployments: Load an existing Agent Spec configuration (JSON), run it on a compatible runtime such as LangGraph and WayFlow, and expose an AG-UI endpoint via an adapter to interact immediately in an AG-UI frontend.
  • Try complex agents quickly: Preview and interact with sophisticated Agent Spec–based agents built by others, or swap in new runtimes without changing the agent definition.
  • Plug-and-play across frameworks: Move assistants between runtimes or swap frontends while preserving behavior and tools; Agent Spec serves as the bridge and AG-UI as the common interface layer.
  • Tracing and observability: With Agent Spec Tracing, agents can emit standardized events to frontends, observability and evaluation services via pluggable hooks.

In short, define once, choose your runtime, and connect to a user-ready UI with minimal effort.

Screenshot of Agent Spec integration in AG-UI interactive Dojo: Users can choose between different runtimes (e.g. WayFlow, LangGraph). The main panel shows a chat-style UI where a weather box widget is generated in the conversation with the user

Get started

Bring your Agent Spec JSON, add an AG-UI adapter to expose your agent endpoint, and connect an AG-UI frontend (e.g., built with CopilotKit) to start interacting in minutes. See the docs and examples below to try end-to-end, then customize for your stack.