Komprise combats ‘MCP bloat’ with a universal interface for AI agents to access enterprise data

Artificial intelligence management firm Komprise Inc. has rolled out a new tool designed to simplify the process of AI agents accessing enterprise data. Titled Universal File MCP, the tool provides a single interface for AI agents to query any kind of data source, regardless of its location. MCP, originally developed by Anthropic PBC, has become a crucial component in the AI infrastructure ecosystem, enabling connections to third-party data sources.

However, companies have been amassing multiple MCP servers to connect with their own platforms, leading to what Komprise co-founder and President Krishna Subramanian calls "MCP bloat." This phenomenon results in lower accuracy, slower performance, and increased costs for enterprises. The issue is exacerbated by the vast amounts of unstructured data that AI agents must process, often spanning millions to billions of files across various storage systems.

These files are often unstructured and siloed across different storage platforms, making it challenging to retrieve the necessary data without incurring high costs.

Komprise's Universal File MCP addresses these challenges by right-sizing AI responses, ensuring that only the relevant unstructured data is sent to AI agents. It identifies the necessary data sources and sends only the required information, enriched with context and governed by user-specific access permissions. The tool leverages Komprise's Global Metadatabase to maintain a consistent schema across all storage resources and employs noise filters to eliminate irrelevant data and prevent processing that could drive up token costs.

The Universal File MCP offers benefits across various agentic use cases. For example, a clinician could ask a large language model to find specific pathology images, and the query would automatically filter the data based on KAPPA-enriched context and user permissions, returning the requested images. Similarly, an information security professional could use the tool to quickly discover non-compliant files or search for archived research data filtered by project keywords.

Komprise's tool works with any data source, whether it's managed by Komprise or not. It simplifies and enhances the efficiency of AI's access to unstructured data at scale, building upon the foundation laid by the first wave of MCP servers. Todd Dorsey, a data center analyst, believes Komprise's approach is a natural evolution of the MCP standard, addressing the need to send AI only the files required to answer questions across various storage systems.