Mistral AI has released an upgraded version of Agentic Search designed to solve challenges faced by businesses building AI systems. The new version incorporates a multi-step retrieval loop along with five core tools similar to file system operations, enabling models to actively reason, check retrieved content, optimize search strategies, and compare information across multiple sources. It outperforms traditional RAG in handling complex structured documents. Tests on financial SEC files and government PDFs show it boosts accuracy significantly and reduces latency and token usage. The tool can be deployed in both cloud and on-premise environments to help enterprises unlock value from sensitive data and build precise retrieval systems.