AG-UI releases a stable specification and SDKs for TypeScript, Python, and .NET, replacing custom agent-to-UI wiring with a single bi-directional event stream.

AG-UI, an open protocol for connecting AI agents to user-facing applications, has reached 1.0 with a stable specification backed by JSON Schema. The TypeScript, Python, and .NET SDKs are now generated from that specification.

The problem it solves

Agents behave differently from traditional request-response systems. They stream tokens, call tools mid-response, update shared state, pause for human input, and spawn subagents. Each agent framework currently implements this wiring differently, forcing developers to build custom connections for every combination of agent framework and frontend.

AG-UI replaces that custom plumbing with one bi-directional event stream. The agent emits typed events, and any client renders them.

Getting started

The fastest path is the CLI scaffold:

npx create-ag-ui-app@latest

This generates a frontend (CopilotKit on Next.js or a terminal client) wired to the agent framework you choose. Set your OpenAI API key and run npm run dev.

To speak the protocol directly, install the Python package:

pip install ag-ui-protocol fastapi uvicorn

The README includes a 30-line Python server example that accepts a RunAgentInput, streams events back over HTTP, and requires Python 3.9 or later. A corresponding TypeScript client example (Node.js 20+) uses @ag-ui/client to connect and handle events like onTextMessageContentEvent and onRunFinishedEvent.

Backwards compatibility and adoption

Version 1.0 is backwards compatible in both directions: 0.x agents work with 1.0 clients and vice versa. A migration guide and changelog are available in the documentation.

AG-UI is adopted by Google, Microsoft, Amazon, and Oracle, and built into most agent frameworks. The repository has 21,047 stars and 2,148 forks on GitHub, with 601 open issues.

Where it fits

AG-UI is one of three open agentic protocols. MCP gives agents tools and context, A2A lets agents work with other agents, and AG-UI brings agents to users. A single agent often uses all three. AG-UI can front agents that speak MCP or A2A, and works alongside generative UI specs like A2UI and Open-JSON-UI, which describe what UI an agent wants to show.

Go deeper


Get the latest AI updates in the Tech AI Magazine app
iPhone · Android

Explain it Like I'm 12

AG-UI Makes Connecting AIs Easier

AG-UI is a new way for AI agents to talk to apps people use. It is now at version 1.0, which means it is stable and ready to use. It comes with tools for popular programming languages like TypeScript, Python, and .NET.

Imagine you are in a school project group. Each person has different ways of sharing ideas. Some use drawings, others write notes, and some speak out loud. This can make it hard to understand each other. AG-UI is like a new rule that everyone in the group agrees on. It helps everyone share ideas in the same way.

Before AG-UI, developers had to create their own ways to connect AI agents to apps. This was tricky because each agent worked differently. With AG-UI, there is now one simple way to connect them. It uses something called a bi-directional event stream. This means that information can flow both ways between the agent and the app.

To start using AG-UI, developers can quickly set up a project with a command. This creates a basic app that connects to the chosen agent. They just need to enter their OpenAI API key and run the app.

AG-UI is designed to work with older versions too. This means it can connect to both new and old agents easily. Big companies like Google and Microsoft are already using it.

AG-UI is one of three main ways for agents to work together. It helps agents connect with users while other protocols help agents work with tools and with each other.

For more details, developers can check out the project on GitHub.