Key takeaways

  • AG-UI is an open protocol that standardises how an AI agent talks to a user-facing application.
  • It works over any event transport, including SSE, WebSockets and webhooks, so you are not tied to one setup.
  • Features include streaming chat, two-way state sync, generative UI and human-in-the-loop collaboration.
  • Many agent frameworks are listed as integrations, from Microsoft Agent Framework to community SDKs.
  • The catches: some integrations are still in progress, and the project grew out of CopilotKit's work.

What it is

AG-UI is an open, lightweight, event-based protocol that standardises how AI agents connect to user-facing applications. It is open source under the MIT licence, and the code is on GitHub at ag-ui-protocol/ag-ui.

The problem it solves

Every team that puts an agent behind a chat box or a dashboard ends up inventing its own way to pass messages, progress and state to the screen. The README places AG-UI alongside other agent protocols: MCP gives agents tools, while AG-UI brings agents into user-facing applications.

The aim is that a front end and an agent backend from different makers can still understand each other, because event formats are matched loosely.

What you get

  • Freedom of plumbing: it works with any event transport, such as SSE, WebSockets and webhooks.
  • Responsive interfaces: real-time agentic chat with streaming.
  • Shared context: bi-directional state synchronisation between agent and app.
  • Richer screens: generative UI and structured messages.
  • A person in the loop: human-in-the-loop collaboration, plus a reference implementation to help teams start.

The catches

  • Maturity: some integrations and SDKs are marked as in progress rather than supported, so check your framework and language first.
  • Support varies: several SDKs are labelled community integrations rather than first party.
  • Provenance: AG-UI was born from CopilotKit's partnership with LangChain and CrewAI, and many of the linked guides come from that ecosystem.
  • Direction: the project points to a public roadmap, which means features are still being built.

What we did not test

We did not run AG-UI, build an agent with it or try any of the listed integrations. We also did not assess its security or privacy behaviour. Every claim here comes from a literal quote in the project's README or the GitHub API, and the feature descriptions are the project's own.

Verdict

Editors' view: this is for developers who are wiring an agent into a real interface and would rather adopt a shared convention than invent one. If your framework is listed as supported, it deserves a look. If it is still in progress, wait or contribute.

Repository
ag-ui-protocol/ag-ui
TypeScript · MIT
View on GitHub ↗

Sources

Need help adopting this, or have a software project in mind?

Tell us what you are building. The initial consultation is free. Email [email protected] or follow @zolnaltd.

Start a conversation →