Building Enterprise‑Ready AI Agents: From Prototypes to Production

Getting an agent to run once is easy, but building one that engineering teams, security reviewers, and auditors can actually rely on is a very different craft. This Workshop helps you move beyond flashy demos to focus on the architecture, controls, and operating discipline required to make agentic workflows dependable in real enterprise settings. Using Claude Code as the development environment and the open-source project as a reference, we'll take a practical look at what a production-grade agent looks like under the hood. We'll explore explicit LangGraph state machines, small named nodes, and the essential plumbing that turns a prototype into a deployable system. We'll cover the design principles behind enterprise-ready agents, including structured atomic workflows and human approval gates for high-risk actions. From there, we'll move into the operational stack, examining the OWASP LLM Top 10 threat model and RAG architecture patterns. We'll conclude with a hands-on blueprint exercise to design a state graph, tool permissions, and monitoring approach—providing a build-ready plan engineering teams can execute immediately.

To fully participate in the interactive components of this workshop, you will need to sign up for accounts at GitHub and at Claude Code (Pro or higher).