Agentic Search for Context Engineering (leoniemonigatti.com)

In a recent workshop titled "Agentic Search for Context Engineering" at AI Engineer Europe 2026, a significant advancement was announced regarding how artificial agents can better curate context for response generation. The speaker emphasized that context engineering—deciding which information to pull from various sources for agents—is crucial for optimizing responses from large language models (LLMs), and argued that effective context curation relies heavily on agentic search, which he estimates accounts for about 80% of this process. This evolution from traditional Retrieval-Augmented Generation (RAG) to agentic RAG marks a critical shift, allowing agents to dynamically determine when and how to retrieve information. The workshop highlighted the increasing complexity of context sources available—ranging from local files to databases, web searches, and long-term memory—along with various search tools, including a versatile "shell tool" capable of executing commands across these contexts. The speaker noted the importance of refining search techniques to enhance accuracy and efficiency, ultimately pointing out that successful agentic search requires thoughtful curation of the search tool stack to meet varying latency and quality requirements.