Exploring the Practical Potential of Model Context Protocols in Software Development

Bachelor & Master Thesis

Model Context Protocol (MCP) has recently gained attention as a promising approach to manage and structure the external context provided to large language models (LLMs) during software engineering tasks. However, it remains unclear whether MCP can effectively address real-world software development challenges such as complex multi-file code generation, long-horizon bug fixing, and consistent state tracking across iterative development sessions. In this project, students will gain hands-on experience with designing and evaluating MCP-based systems, including learning how to decompose software tasks into structured context units, implement retrieval mechanisms for relevant context slices, and assess the impact of context selection on model performance. They will also explore trade-offs between context completeness, granularity, and retrieval efficiency, and develop a deeper understanding of how structured context influences LLM behavior in practical software engineering workflows.

Required knowledge:

• Strong programming background, especially proficient in python.
• Familiar with static analysis techniques.