This paper presents an anomaly detection approach tailored for distributed microservice architectures, leveraging data streaming through Apache Kafka and advanced machine learning techniques for real-time performance analysis. The method incorporates root cause localization to precisely identify the sources of anomalies within complex service interactions. Benchmarking results demonstrate the approach’s ability to detect a broad range of anomalies, including latency spikes, service crashes, and resource contention, with high accuracy and low false positive rates. The system’s scalability is validated through performance evaluations in large-scale microservice environments, with results showing effective anomaly detection across systems with up to hundreds of services. This paper also discusses the selection of Key Performance Indicators (KPIs), which are chosen based on their relevance to system performance and their ability to indicate service degradation or failure. The proposed method offers a significant improvement in real-time anomaly detection, providing both actionable insights and a scalable solution for complex architectures.