CopilotKit, an open-source company building infrastructure for AI-powered software agents, said it raised $27 million in Series A funding and launched an enterprise product aimed at helping companies build user-facing “agentic” applications.
The round was led by Glilot Capital, NFX and SignalFire, the company said.
CopilotKit also announced CopilotKit Enterprise Intelligence, a self-hostable product designed to help enterprises deploy AI agents inside applications used by employees and customers. The company said its tools are used by a majority of Fortune 500 companies and counts Deutsche Telekom, Docusign, Cisco and S&P Global among its customers.
CopilotKit is the company behind AG-UI, an agent-user interaction protocol designed to connect AI agents with human users inside software applications. The company described AG-UI as part of an emerging “Agentic Protocol Stack,” alongside MCP, which connects agents to external tools, and A2A, which connects agents to other agents. AG-UI is intended to handle interactions between agents and people.
The company said AG-UI has been adopted by Google, Microsoft, Amazon and Oracle, as well as AI frameworks including LangChain, Mastra, PydanticAI, Agno and AG2.
CopilotKit said its technology powers tens of millions of agent-user interactions each week. It also said its open-source repositories have driven millions of weekly installs and have received more than 40,000 GitHub stars.
The company said the software industry is shifting away from traditional software-as-a-service interfaces toward a model built around APIs for agents and generative interfaces for people. In that model, AI agents move beyond chatbots and into applications themselves, including dashboards that change dynamically, voice agents, mobile apps that adapt to data and Slack workspaces where humans and agents collaborate in real time.
“As agents move into production, companies are discovering that the old patterns for building software no longer hold,” the company said. “Agents do not behave like traditional software: they stream, wait, resume, collaborate with humans and persist across sessions, devices and long-running workflows.”
CopilotKit said it is building what it calls the “agent-user interaction stack,” a layer between agent backends and the applications people use.
“Every major software category is being rebuilt around agents,” said Atai Barkai, co-founder and CEO of CopilotKit. “The leaders are learning what it takes to turn an agent into a real product experience. The hard part is not generating text. It is building the interaction layer — the layer that makes these systems usable, stateful and deeply integrated into software. That’s the category we are building.”
CopilotKit Enterprise Intelligence is designed to let companies integrate AG-UI-compatible agents into user-facing software. The company said the product supports generative user interfaces, MCP apps, multimodal file uploads, voice transcripts and continuity across sessions, workflows, reconnections and mobile conditions.
CopilotKit said the product is being expanded to include an insights layer, described as user analytics for agentic products, and a self-learning layer that allows agents to improve through real user interactions.
For enterprises, the company said, the product is meant to help teams put agentic products into production while keeping systems self-hosted on customers’ existing cloud infrastructure for security, control and data boundaries.
“Enterprise developers do not want another closed agent product,” said Uli Barkai, co-founder and head of growth at CopilotKit. “They want the standard way to build agentic applications on top of whatever backend and cloud stack they already chose. That is what CopilotKit has become in open source, and this funding lets us scale that foundation further.”
Investors said the company is positioned to become a core infrastructure layer for AI software.
“When we first met Atai and Uli, we were struck not just by the strength of the technology and vision of CopilotKit, but also by the ecosystem’s response,” said Lior Litwak, managing partner at Glilot Capital and head of Glilot+. “At Glilot Capital, we’ve spent over a decade investing in the infrastructure layers that define how enterprises build and secure software. CopilotKit represents the next one: the standard interface between humans and AI agents.”
Litwak said the speed at which cloud providers and open-source frameworks have converged around CopilotKit and AG-UI shows the category is forming.
Gigi Levy-Weiss, general partner at NFX, said enterprise adoption of agentic software has been slowed by a lack of tools.
“Everything is going to be rebuilt around agents. But right now, there’s a critical gap: enterprise adoption has been slow because the tooling hasn’t kept pace,” Levy-Weiss said. “CopilotKit is building the front-end stack that closes it — and the developer community has already responded.”
CopilotKit is headquartered in Seattle.