Automotive software development requires high reliability, safety standards, and performance, with C++ being the primary programming language. This thesis examines the effectiveness of multi-agent Large Language Models (LLM) for automated C++ code generation and testing. This study proposes CPPCoderAgent, a framework that uses LLM agents for cooperative C++ software development. The system consists of a Programmer Agent and a Test Designer Agent. The agents collaborate through iterative feedback loops, in which the Test Designer Agent generates test cases, the system executes the tests, and the resulting feedback assists the Programmer Agent in code revision. The results show that incorporating the Test Designer Agent improves code accuracy compared to using the Programmer Agent alone. However, the analysis revealed limitations, including LLM hallucination and incorrect test case generation, that affect the system’s trustworthiness. These findings indicate that although multi-agent LLM systems demonstrate potential for improving code generation efficiency, they cannot yet replace human supervision in safety-critical automotive applications. In summary, this research contributes to the understanding of both the capabilities and limitations of LLM-based multi-agent systems in automotive software development.