1.2. The Foundry Mental Model
💡 First Principle: Foundry has a strict containment hierarchy — resource → project → deployments/agents/connections — and almost every "where does this live / what do I create first" question is answered by knowing that hierarchy. The platform groups work (projects) inside billing-and-identity boundaries (resources), and puts the things you actually use (deployed models, agents, data connections) inside projects.
Picture a company: the resource is the building (one address, one lease, one security desk — billing and identity live here). A project is a team's floor (its own workspace, its own connected tools, but sharing the building's security). A deployment is a specific machine that team installed on their floor (a particular model, at a particular version, with a quota). An agent is a worker on that floor who uses those machines and the team's filing cabinets (connections) to get jobs done. You don't put a machine in the lobby; you don't give a worker a desk in the parking lot. The hierarchy is the answer to most placement questions.
⚠️ Common Misconception: Learners often flatten this and imagine "a model endpoint you stand up," like a web service. In Foundry you don't build the model endpoint — you deploy a catalog model into a project, and the platform provides the endpoint. Getting the hierarchy backwards leads to wrong answers about what to provision and in what order.
