Roberto Stagi

Roberto Stagi is co-founder and chief technology officer of Ratel, an open-source infrastructure company addressing a practical limitation of AI agents: expanding their tools and instructions can make them slower, costlier, and less reliable. His work focuses on context engineering: providing agents with the information and capabilities relevant to their immediate task.

His engineering career included consumer-product consulting at Reply, technical leadership at Mia Platform, and serving as chief technology officer of SavingBud. At Bending Spoons, he worked on Remini and Evernote before leaving to build with fellow founder Giacomo Nicoli.

Their initial product, Agentified, aimed to embed agents inside existing applications by connecting interfaces, authentication, and product functionality. They subsequently shifted toward the infrastructure challenge behind those integrations: managing growing inventories of tools, instructions, and context without degrading agent performance. That work became Ratel, whose open-source implementation uses selective tool retrieval, including BM25 and semantic search, to expose relevant capabilities progressively instead of loading everything into an agent’s prompt.

• Context is production infrastructure. Stagi treats irrelevant tools, stale instructions, and bloated prompts as operational problems that affect reliability and cost. His account of building Ratel connects those recurring failures to the company’s founding.

• TypeScript belongs in the agent application stack. A longtime JavaScript developer who adopted TypeScript, Stagi argues that Python remains well suited to training, research, and GPU serving, while production-facing agents can benefit from sharing a language with their surrounding applications. His AI Engineer World’s Fair presentation emphasizes npm’s existing application infrastructure and the architectural benefits of keeping agent loops, tools, backend services, and interfaces together.

• Shared schemas reduce integration boundaries. Stagi advocates end-to-end type safety across model-facing data, application logic, and interfaces, allowing one schema to travel through the stack instead of synchronizing separate service contracts.

Stagi also organizes AI Builders Milan and helped bring AI Socratic to Milan, cultivating technically focused discussions among engineers, founders, and researchers.