MLE-Agent is a multi-model AI agent designed for ML engineers and researchers. It integrates with Arxiv and Papers with Code to access state-of-the-art methods and build autonomous baselines. The framework supports OpenAI, Anthropic, Gemini, and Ollama providers, features Code RAG for context-aware assistance, and includes smart debugging with automatic debugger-coder interactions. It can independently participate in Kaggle competitions and complete end-to-end ML tasks.