AI agents are moving from experimental demos to real, production-ready systems. But as soon as we go beyond a single agent, things get messy: coordination, task delegation, memory, feedback loops, and reproducibility all become hard problems.

This talk introduces mini007, an R framework designed to simplify the creation, management, and orchestration of AI agents. In a 20-minute, practical session, we will explore how mini007 models agents as modular, goal-driven components that can collaborate, specialize, and evolve within a controlled workflow. We will cover the core concepts behind agent design, roles, tools, communication, and autonomy, and show how mini007 helps structure multi-agent systems without sacrificing transparency or control.

Through concise examples, the audience will see how to define agents, coordinate their interactions, and manage complex tasks such as analysis, planning, and execution.The talk emphasizes engineering principles over hype: reproducibility, observability, and composability. Whether you are experimenting with AI agents, building decision-support systems, or exploring agentic workflows in data science, this session will provide a clear mental model and concrete tools to get started.