A TypeScript SDK for building agent-native apps with generative UI, shared state, and human-in-the-loop flows in React, Angular, Vue, and chat apps.
ag-ui-protocol/ag-ui
An open, event-based protocol that standardizes how AI agents connect to user-facing frontend applications.
- Stars
- 16,716
- Forks
- 1,728
- Open issues
- 495
- Last commit
- 8 hours ago
What is it?
AG-UI, short for Agent-User Interaction Protocol, defines a lightweight event format that AI agent backends emit during execution, so frontend applications can display agent output and respond to it in a consistent way. Agents accept a small set of AG-UI compatible inputs, and the protocol is transport-agnostic, so it can run over SSE, WebSockets, webhooks or other channels.
The project positions itself alongside MCP, which gives agents tools, and A2A, which lets agents talk to each other: AG-UI covers the link between agents and the user interface. It supports real-time streaming chat, bi-directional state synchronization, generative UI, frontend tool calls and human-in-the-loop interaction. A reference HTTP implementation and a default connector are included.
The repository lists integrations with a range of agent frameworks, including LangChain, CrewAI, Google ADK, Pydantic AI, LlamaIndex and Mastra, plus client SDKs in several languages and a TypeScript core package. A CLI, npx create-ag-ui-app, scaffolds a new application.
Key features
- 01Streams agent output to frontends in real time through standard event types
- 02Synchronizes state bi-directionally between agents and the user interface
- 03Supports generative UI and structured messages
- 04Lets agents call frontend tools and request human input
- 05Works over any transport such as SSE or WebSockets
Use cases
- Adding a chat interface with streaming responses to an existing agent backend
- Letting an agent read and update application state shown in a web or mobile UI
- Building human-in-the-loop approval flows where users confirm agent actions
- Connecting agents built with LangChain, CrewAI or Pydantic AI to a shared frontend protocol
Quick start
- 1
Create a new AG-UI application with the create script
npx create-ag-ui-app my-agent-app
Things to know
- The protocol is still evolving, and some framework integrations are marked as in progress.
- Using it well typically means wiring up an agent backend, so it is not a standalone tool.
Topics
FAQ
- What is ag-ui?
- An open, event-based protocol that standardizes how AI agents connect to user-facing frontend applications. AG-UI, short for Agent-User Interaction Protocol, defines a lightweight event format that AI agent backends emit during execution, so frontend applications can display agent output and respond to it in a consistent way. Agents accept a small set of AG-UI compatible inputs, and the protocol is transport-agnostic, so it can run over SSE, WebSockets, webhooks or other channels.
- Is ag-ui free to use? What license does it have?
- ag-ui is open source under the MIT license. Check the license terms before using it in commercial products.
- Is ag-ui still maintained?
- Actively maintained. The last commit to ag-ui was on Oct 20, 2026.
- What can I use ag-ui for?
- Adding a chat interface with streaming responses to an existing agent backend Letting an agent read and update application state shown in a web or mobile UI Building human-in-the-loop approval flows where users confirm agent actions Connecting agents built with LangChain, CrewAI or Pydantic AI to a shared frontend protocol
- How do I install ag-ui?
- Create a new AG-UI application with the create script: npx create-ag-ui-app my-agent-app
- Where can I get ag-ui?
- The source code and installation instructions are on GitHub: https://github.com/ag-ui-protocol/ag-ui
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A framework-agnostic Generative UI toolkit built around OpenUI Lang, a compact streaming language that lets LLMs produce renderable UI from your own component library.
Self-hosted AI coworker platform where each agent gets its own browser, files and tools, with every action gated by policy and logged.
An open-source chat interface where AI answers can be interactive, rendering explanations, comparisons, charts, and working tools built with CopilotKit and AG-UI.
TypeScript framework for building AI agents with memory, tools, MCP, RAG, voice and workflows, plus the VoltOps console for observability.
A TypeScript framework for building MCP servers, MCP Apps with React views, and clients, with a built-in Inspector and CLI.