The updates, arriving this October, address the friction between existing data architecture and the requirements of autonomous agents. Prakash Darji, General Manager of Data & Digital Experience at Everpure, noted that enterprise AI often stalls not due to model limitations, but because data remains unprepared for real-time interaction. The new tools integrate the open Model Context Protocol (MCP), allowing AI agents to query live data catalogs through natural language without requiring custom API development.

Performance enhancements target the hardware-software bottleneck directly. The introduction of PureKVA (Key-Value Accelerator) for FlashBlade systems pre-stages context into GPU memory, claiming a 20x improvement in Time to First Token. Alongside this, the platform introduces DeepReduce, a compression engine that identifies sub-block similarities to expand usable storage capacity automatically. These features are designed to function without data relocation, keeping information secure within its primary system of record while lowering the overall costs associated with external API token usage.