AG-UI brings agents into user-facing apps: about 16 standard event types unify both ends and work over any event transport
Synopsis
AG-UI is an open, lightweight, event-based protocol that standardizes how AI agents connect to user-facing applications. Previously there was no shared convention for wiring an agent into an interface; AG-UI has agent backends emit events matching one of about 16 standard event types during execution and accept a few simple inputs, while a middleware layer adapts to any event transport such as SSE, WebSockets or webhooks. It grew out of CopilotKit's early partnership with LangChain and CrewAI, ships under the MIT license, and comes with a reference HTTP implementation and a default connector.
Interpretation
Agent-to-interface work now has a public event convention: events an agent backend emits during execution are accepted by a front-end application as long as they match one of AG-UI's roughly 16 standard event types. Previously there was no such shared event convention for connecting agents to application interfaces; AG-UI separates this layer and places it alongside MCP, which gives agents tools, and A2A, which lets agents talk to each other. The protocol is released as open source under the MIT license and ships with a reference HTTP implementation and a default connector so teams can start quickly.
A middleware layer keeps the protocol compatible across environments, and loose event-format matching lets many agents and apps interoperate. The middleware layer is not tied to one transport; SSE, WebSockets, webhooks and other event transports can all carry AG-UI events. The AG-UI Dojo demonstrates the protocol's core building blocks with simple examples of 50 to 200 lines of code each, and the Dojo and framework integrations are published as source code.
Perspective
The protocol targets developers connecting AI agents to user-facing applications; the material provides a protocol description, a reference HTTP implementation, a default connector and Dojo examples of 50 to 200 lines of code each, with no quantitative performance or interoperability results.
The material positions AG-UI as a third agent protocol complementary to MCP and A2A, with compatibility achieved through loose event-format matching; it gives no actual cross-framework interoperability test results.