Komprise launches Universal File MCP for AI data access

Komprise has launched Universal File MCP, a product designed to give enterprises AI-driven access to unstructured data.

The software provides a single interface for AI agents and large language models to query file storage, network-attached storage, cloud repositories and other unstructured data silos.

Komprise is addressing a problem that has emerged as more suppliers publish Model Context Protocol interfaces for AI tools. It argues that multiple tool definitions and very large file and object listings can overwhelm AI systems with irrelevant information, increasing token usage and reducing answer quality.

The issue is becoming more visible as companies try to connect AI models to large stores of corporate data. Komprise cited McKinsey research showing that 60% of agentic AI computing costs go to response refinement, while a separate Dun & Bradstreet survey found that only 5% of enterprises consider their data ready for AI.

How it works

Universal File MCP is designed to limit the amount of data that reaches an AI model by filtering access based on user permissions and metadata. The system returns only the data a user is authorised to see and keeps an audit trail of what is sent to an AI tool.

The product also uses the company's Global Metadatabase to create a common schema across different storage systems. That is intended to give AI tools a more consistent way to query unstructured data, regardless of where the files are stored or who created them.

Komprise says its AI Preparation & Process Automation tools can add contextual metadata based on industry, sensitivity and user context. Its analytics tools can also remove outdated, conflicting, irrelevant and duplicate data before queries are passed on.

Another feature is what Komprise describes as progressive disclosure. In practice, an AI tool can start with metadata for summarisation and retrieve files later only if needed, with results exportable as an Apache Iceberg table.

The software is also intended to work with tiered and archived data. According to Komprise, data remains accessible even after being moved between storage tiers, and query results can then be fed into AI systems, analytics platforms or lakehouses for further processing.

Use cases

Komprise outlined several use cases for the product, including enterprise search, support for agentic AI workflows and IT diagnostics. In healthcare, a clinician or researcher could use a large language model to search pathology images linked to specific studies and time periods, with the query filtered through metadata and access controls.

In legal or commercial settings, AI agents could identify relevant liability clauses across customer contracts and use those results to draft a new template. IT teams could also use the system to identify non-compliant files, isolate data for review or search for ageing research data that may be suitable for tiering.

"The Komprise Universal File MCP lifts the veil on dark enterprise data by unifying governed access across silos," said Kumar K. Goswami, Chief Executive Officer, Komprise.

"We're thrilled to help our customers easily leverage AI across all of their file data by asking complex questions and initiating workflows to act on the results, all while ensuring access controls remain intact," Goswami said.

Children's Medical Centre Dallas is among the organisations Komprise cites as a use case for the technology.

"Opening up clinical and research data to AI tools means my team needs to review who could see what, case by case, putting security in the middle of every request," said Stephen Clark, Director of Information Security, Children's Medical Centre Dallas.

"With Komprise Universal File MCP, the query respects the permissions a user already has and leaves us a record of what went to the model. Our researchers get their data faster, and IT is not managing a separate connector for every storage system," Clark said.

Industry analyst Todd Dorsey said the broader challenge is no longer just connecting AI to enterprise data, but controlling the volume and relevance of the material returned.

"The first wave of MCP servers solved the issue of connecting AI to enterprise data," said Todd Dorsey, Senior Storage Analyst, DCIG.

"But it's critical to send AI only the files it needs to answer questions across the mix of NAS, object and cloud storage enterprises typically run rather than a single vendor footprint. Komprise is taking a notable step toward simplifying the curation of unstructured data for AI with its new MCP solution," Dorsey said.

Komprise said Universal File MCP is available through an early access programme for customers and partners.