In humans, mice, and other mammals key internal organs such as the gut, the lungs, the pancreas, and the liver all derive from the same embryonic tissue: the endoderm. The development of all of these structures thus depends on a same set of early cells, and on the developmental instructions provided to them by the embryo. One approach to study what these instructions might be is to study the behaviour of endoderm cells when they can not rely on embryonic instructions. That is, when they are generated and made to develop outside of the embryo. The rationale behind this approach, widely embraced by the field of synthetic embryology, is that the unguided development of specific embryonic cell types can provide a window to the observation of intrinsic characteristics of these cells and of fundamental developmental principles which might not be executed or manifested within the regulated environment of the embryo. Accordingly, in the work here described I use gastruloids, synthetic models of development made by aggregation of mouse embryonic stem cells in vitro to characterise the developmental behaviour of endoderm cells, in the absence of embryonic cues and in the absence of the extraembryonic tissues that they would usually rely on in vivo. In the first chapter of this thesis, I provide an introduction, overview, and review of our current understanding of the developmental biology of mouse endoderm as this develops in the embryo. I also introduce gastruloids as stem-cell-based experimental systems to study developmental mechanisms in vitro, decoupled from the constraints of experimental work with animals and animal embryos. Chapter 2 is dedicated to the earliest steps of endoderm development in gastruloids. I show when and where these cells appear, and how they sort as the gastruloid starts to mature. I find that endoderm cells follow similar spatial and temporal developmental patterns as those they would follow if they were in the embryo. I notably find that they always remain in an epithelial state, even as other cell types undergo mesenchymal transition. In Chapter 3 I describe surprising dynamics of endoderm cells over time, and how they self-organise into complex architectures at the core of the gastruloids. The epithelial domain that forms is dynamic and plastic, and partitions the gastruloids into interfacing compartments. Importantly, I reveal a significant endodermal component to this in vitro system, usually considered to be unstructured and neural/mesodermal in character. In Chapter 4 I show how the core endoderm compartment that I see forming in gastruloids is patterned along the longitudinal axis of the aggregate, with different cellular identities at different anterior-posterior position of the domain. By comparison with published embryonic datasets, I identify endodermal identities corresponding to the full spectrum of those that make up the gut tube in the embryo. Specifically, I describe how most endodermal cells adopt an early intestinal fate, and how many acquire anterior foregut identities. The work described in this thesis presents gastruloids as promising in vitro systems for the study of endoderm biology, morphogenesis, and mature fate specification, filling a gap within the field of synthetic embryology. It also notably opens translational possibilities to the use of gastruloids to produce endodermal cell types which may be otherwise difficult to generate using conventional directed differentiation methods.