Earlier this year, a new kind of AI-powered agent called OpenClaw drew widespread attention across the tech world for its ability to connect with the entirety of a user’s digital world — email, apps, tools and workflows — and autonomously perform tasks in the background.
Now, the University of Michigan is deepening its role with that emerging “agentic AI” technology with the launch of the Institute for Agentic Computing.
The institute will be based in the Office of the Vice President for Research and serve researchers and developers who seek to build and apply agentic AI infrastructure to advance scientific discovery, engineering and the broad knowledge economy. It will also initially be a central node for managing developers and maintainers of the OpenClaw platform.
Agentic AI is a type of artificial intelligence that can accomplish specific goals with limited supervision. A cybersecurity agent might continuously monitor a company’s network for vulnerabilities, while software development agents might work in concert to write code, review it, generate tests and fix bugs.
One example of this is OpenClaw, created by Austrian web developer Peter Steinberger and recognized by its cartoon lobster logo. It is a free, self-hosted AI assistant that runs on a computer or server and acts like a proactive digital agent. OpenClaw can connect AI models like OpenAI’s ChatGPT to a user’s everyday apps to automate tasks, manage files and organize a user’s digital life using customizable “skills.”
“The future of AI is clearly agentic, moving from interactions with simple chatbots to networks of agentic systems collaborating with humans and, importantly, other AI agents to solve problems,” said Karthik Duraisamy, director of the Michigan Institute for Computational Discovery and Engineering.
“Frameworks like OpenClaw are enabling an exciting entry into this agentic world, and the University of Michigan is well positioned to play a central role in its evolution and application by advancing the technology, serving as a key node in the global agentic ecosystem, and contributing to nationally significant efforts like the Genesis mission for AI-enabled science.”
The institute’s leadership team includes Duraisamy and Kurt Skifstad, the Dixon and Carol Doll Executive Director of the Center for Entrepreneurship in the College of Engineering.
Beyond its role as a research and development hub, the institute will also address foundational challenges of governance, alignment, trust and responsible development that this technology demands.
“The institute exists to make sure agentic AI systems are built openly, responsibly and in service of problems that matter,” Skifstad said.
The Institute for Agentic Computing builds upon a firm foundation at U-M developed by Information and Technology Services over the last two years under the leadership of Vice President for Information Technology and Chief Information Officer Ravi Pendse, said Daniel Feder, senior managing director of investments in the U-M Investment Office.
“It’s exciting to see where this technology has taken us and where we’ll continue to go from here,” said Feder, adding that the university also sees significant investment potential in the technology.
As part of the partnership, the university’s endowment is a founding investor in the Lobster Compute Company, a newly formed company designed to commercialize and expand agentic computing technologies. LCC will also use a portion of its capital to support the OpenClaw community, helping to position LCC as a trusted partner for teams seeking to build important companies using OpenClaw’s open-source toolset.
“This is a way for us to invest in companies that will be built upon the OpenClaw platform,” Feder said.
The partnership formally kicked off April 16 at “ClawCon Michigan,” a free event that brought together developers and AI enthusiasts at the Crisler Center to network, see demonstrations of personal AI tools and workflows and discuss issues related to agentic AI.
