Why We're Watching This: AG-UI adoption across Google, Microsoft, Amazon, and Oracle signals it could become the default standard for agent UI communication. If it does, the tooling layer built on top of it becomes critical infrastructure for every team shipping agentic products.
Key Takeaways
- CopilotKit has raised $27 million in Series A funding led by Glilot Capital, NFX, and SignalFire to build enterprise tooling on top of its open source AG-UI protocol
- AG-UI standardises how AI agents connect to and communicate with user interfaces, enabling streaming chat, front-end tool calls, and state sharing inside apps
- The protocol is supported by Google, Microsoft, Amazon, Oracle, LangChain, Mastra, PydanticAI, and Agno, and sees millions of installs per week
- Enterprise customers include Deutsche Telekom, DocuSign, Cisco, and S&P Global, with a large portion of Fortune 500 companies using AG-UI in production
- CopilotKit is launching CopilotKit Enterprise Intelligence, a self-hostable bundle for fully deploying agents within apps.
CopilotKit has raised $27 million in Series A funding to commercialise its open source AG-UI protocol, which standardises how AI agents connect to and communicate with user interfaces inside applications.
The protocol works alongside the widely adopted Model Context Protocol (MCP) and Google’s Agent2Agent (A2A) protocol, giving developers a framework to deploy agents that can render dynamic, interactive UIs rather than returning blocks of text.
Instead of a user asking a travel app to plan an itinerary and receiving a paragraph of output, an agent using AG-UI can return a structured, interactive interface built from the company’s own component library.
“The agent can reply to you, not just with blocks of text, but with interactive UIs that are defined by your own company,” CEO Atai Barkai told TechCrunch. “If a user asks for a breakdown of revenue by category, instead of getting a big, impenetrable paragraph, you get a pie chart — your own design of the pie chart that the user can interact with.”
CopilotKit now supports LangGraph JS and LangGraph Platform 🤝 You can now build agentic applications with shared state, customizable UI, and human-in-the-loop capabilities.
Read below to see how you can debug your agent interactions and deploy at scale with LangGraph Platform’s… pic.twitter.com/svtQpCG2RW
— LangChain (@LangChain) December 10, 2024
Enterprises want optionality and they want self-hosting. These are two things they don’t really get in the Vercel stack. — Atai Barkai, CEO, CopilotKit
AG-UI is already supported by major infrastructure providers including Google, Microsoft, Amazon, and Oracle, as well as popular agent frameworks like LangChain, Mastra, PydanticAI, and Agno. The protocol sees millions of installs per week and is in production use across a large portion of Fortune 500 companies, according to the company.
The commercial product, CopilotKit Enterprise Intelligence, is a self-hostable offering layered on top of AG-UI that adds enterprise support, deployment infrastructure, and the controls businesses need to run agents inside their own products.
The company positions this as complementary to, not a replacement of, the open source protocol — a pattern mirroring how enterprise agent deployment timelines are compressing across the industry.
AG-UI works alongside MCP and A2A rather than competing with them. Engineering teams evaluating agent infrastructure should assess all three protocols together, as they address different layers of the agent communication stack.
CopilotKit competes with Vercel’s AI SDK, assistant-ui, and OpenAI’s Apps SDK, though it argues its horizontal, framework-agnostic approach gives enterprises the optionality and self-hosting control that vertically integrated alternatives do not.
The company currently has 25 employees and plans to use the new funding to grow its team.
Most enterprise AI projects stall not because the model fails, but because the output lands in a chat box and stops there. AG-UI and frameworks like CopilotKit exist to close the gap between what agents can do and what users can actually see and interact with.
Engineering teams building production agent workflows should evaluate their UI layer before the model layer, not after. Structured comparisons of agent tooling are available on Relve, an AI tools intelligence platform.
