This Bachelor’s Thesis addresses the challenges faced by developers when working with files in microservices environments. Given the large number of services and configurations involved, a minor error can unexpectedly block the entire system. The aim of the project is to design and implement a complete system based on artificial intelligence agents, capable of detecting and correcting errors through the application of Retrieval-Augmented Generation (RAG) techniques. This approach allows access to both an internal knowledge base (Chroma DB) and external sources (Serper), avoiding hallucinations and returning information grounded in verified data. The proposed solution relies on a modular architecture based on the n8n orchestra- tor, the Streamlit interface, and the Chroma DB vector knowledge base, complemented with external searches using Serper and Large Language Models (LLMs) accessed through OpenRouter. For the system evaluation, several real and synthetic datasets were used, including configuration files from different services, Docker files, and Python code. The validation considered not only the system’s ability to correct syntactic, semantic, and logical errors, but also its ability to meet functional requirements expressed as questions. The results indi- cate that the implementation of RAG strategies significantly improves system performance, achieving acceptable success rates that could be further increased by using more power- ful models. Furthermore, it was observed that the most expensive model does not always guaranty the best performance; in one multi-error scenario, an intermediate configuration outperformed the most powerful model, suggesting that the most expensive model is not always the best choice.s