A Comparative Study of Model-First API Design Frameworks and a Methodology for Endpoint Interdependency Analysis

Abstract

In an era where software systems form the backbone of digital infrastructure, Application Programming Interfaces (APIs) play a crucial role in enabling interoperability between services, applications, and users. APIs define structured contracts, sets of endpoints that expose specific functionalities of a system. As modern architectures increasingly rely on distributed and cloud-native services, the consistency, scalability, and maintainability of APIs have become vital for efficient system design and evolution. This thesis focuses on the Model-First API Design approach, where the API contract is defined before implementation, using formal modeling frameworks. Three such frameworks are examined: gRPC/Protocol Buffers, AWS Smithy, and Microsoft TypeSpec. Each framework provides a structured way to describe endpoints, operations, and data models, supporting automated documentation and SDK generation. Building on this foundation, the thesis introduces a methodology for static inter-endpoint dependency analysis that leverages the structured models produced by these frameworks. The analysis identifies data dependencies between API endpoints by comparing shared types and message structures in their input and output models. This enables the visualization of relationships between operations and the discovery of implicit invocation orders that describe real API workflows. Experimental results show that Smithy and TypeSpec generate highly accurate and noise-free dependency graphs, effectively capturing object-level dependencies and reducing ambiguity. In contrast, gRPC exhibits reduced precision due to its inability to wrap primitive types, which leads to semantic loss in its OpenAPI export. The findings demonstrate that well-structured, model-first definitions provide a richer and more reliable foundation for analyzing API interdependencies than traditional static or dynamic approaches such as those implemented in RADAR.