FRAME

AI agents powered by large AI models automate interactions between humans and digital tools and are increasingly adopted to enhance productivity and integrate local and external data sources and services.

However, current agent technology is still unable to provide the reliability, transparency and composability required for scalable and trustworthy deployment. The lack of engineering foundations and systematic quality control prevent agents from evolving into trustworthy, reliable, composable and maintainable systems.

FRAME aims at advancing the engineering of AI agents by creating a software-engineering foundry layered on top of existing agent technologies.

The foundry will provide reference architectures, model-driven engineering methods, lifecycle governance, quality control and selfimprovement to support the development of modular, composable and trustworthy agents able to reason about uncertainty. It will deliver the engineering abstractions, methods and tools needed to support reliable, scalable end-to-end development, deployment and evolution of AI agents. By establishing foundations for trustworthy agent engineering, FRAME will enable productivity, adaptability, reliability and foster both adoption and economic value across sectors. FRAME’s will be validated on real-world use cases (robotics, healthcare and software frameworks) settings and will also advance and disseminate critical knowledge and skills in AI agent engineering to industry professionals, postgraduate students and researchers. This research is vital to the digital and AI-enabled economy and its potential to generate opportunities for productivity and growth. The project will build and share an understanding of trustworthy and safe AI agent engineering and will sustain its benefits to the economy and society.