MCP for Cloud Storage: A Real-World Example From Wasabi and Performance Hub

Performance Hub started as an answer to a specific operational problem. Its founder, Tim West, also runs UBX, a boxing and strength training franchise with more than 80 locations across eight countries. After COVID forced clubs to operate with reduced staff, West's team needed a way to automate access and security across the network without adding headcount at every location. That build became Performance Hub: a platform that now handles video surveillance, door access, computer vision, and a layer of AI agents built on top of it all.

Performance Hub has been a Wasabi customer for some time, storing the video, metadata, and access records that its platform generates across its client base. More recently, it became one of the first to connect to Wasabi's new MCP server during its beta release. That connection turned into something concrete: AI agents querying data in Wasabi directly, and a measurable shift in how Performance Hub operates day-to-day.

What is MCP, and why does it matter for cloud storage?

MCP, or Model Context Protocol, is an open standard that defines how AI agents connect to external systems, including backend storage, through natural language. Wasabi's MCP server gives any AI agent standardized access to Wasabi storage, enabling over 150 tools spanning S3 operations, identity and access management, utilization and usage metrics, and control account management, with no lock-in to any particular AI vendor, model, or framework. A team connects an agent once, over OAuth 2.0, and from there the agent can act on the data directly.

The practical effect is that a person can ask a question in plain language, and the agent decides on its own which of those tools to call to answer it. There’s no need to know the underlying API, the authentication scheme, or how the data is structured. David Nedved, Performance Hub's CTO, describes it this way: "You talk to the data. You talk to Wasabi... When have you ever spoken to a product like you would a human being?" That shift, from writing code against an API to asking a question in the same language you'd use with a colleague, is what MCP is built to support at the protocol level.

How Performance Hub is using MCP

Performance Hub had built its own internal MCP server before Wasabi released the beta. Every API on its platform, billing, camera metadata, door and gate access logs, was already exposed as a tool its own agents could call. That precondition is why the connection worked immediately, with no integration project. Both systems were already built on the same assumption: agents should reach data directly, not through a human-maintained interface or dashboard.

Nedved describes the result: "That existing system has allowed our agent and our platform that was already built to work out of the box with MCP to instantly work overnight with Wasabi's MCP." Performance Hub has used Wasabi's MCP since the beta began, primarily to let its own AI agents query the unstructured data it stores in Wasabi, without routing that work through separate agent software.

One example from Nedved illustrates what this looks like in practice: he asked an agent how much storage a specific franchise location was using. Answering that question required the agent to resolve a location name to Wasabi's internal bucket identifier, then retrieve the storage figures tied to that bucket, entirely through MCP calls, without Nedved specifying how to do either step. "It's actually had to go and query another system because in Wasabi we call our customer buckets their UUIDs," he said. "So it's gone and figured out that. It's mapped that, checked the storage, all using MCP."

The data Performance Hub queries this way is not limited to video. Nedved continued: "Access logs, which are queryable through MCP from our door and gate access also go in Wasabi... It's not just computer vision, right?" Access records, billing data, and camera metadata all sit in the same storage layer, and MCP gives Performance Hub's agents a single way to reach any of it.

The implications for non-technical teams

The more notable shift did not happen in engineering. Tim West, Performance Hub's CEO, is not a developer, and MCP changed what that means for his day-to-day.

"As soon as that MCP was released, my agent had mainline access to the data that I would've otherwise had to ask David to compile for me or ask the tech team to compile for me," West said. "So, from my point of view as an operator, the combination of an MCP with my own agents was a huge unlock."

MCP is often described in terms of what it removes for engineers: hand-built integrations and custom authentication logic. West's experience points to something different: a closed gap between a business leader and his company's data, regardless of whether that leader could write a line of code.

Nedved describes the same shift at a broader scale. "When in your traditional human development history, have you ever picked up a product and asked the product how it works?" he said. "We don't look at documentation, we don't look at APIs. You ask the question and the agent goes and talks to the data and then also talks to the product... This allows ordinary business users to essentially just talk and build a product with natural language."

Traditional dashboards are designed around the questions someone predicted in advance. That design choice has a cost: whatever wasn't predicted goes unanswered. An agent removes that constraint entirely, producing the answer on demand rather than from a fixed report. For an operator like West, the result is that a business decision no longer waits on an engineering queue.

Why MCP’s value compounds over time

The data Performance Hub stores in Wasabi doesn't change once it's written. The models capable of reading it do. Nedved frames the implication directly: "The answer to a question that only sits in Wasabi storage today, may be something I can find and answer on my own tomorrow."

An MCP-accessible dataset doesn't need to be rebuilt every time a better model comes out. The same stored data becomes more useful on its own, as the agents reading it get smarter, without any extra engineering work. That matters because most companies either overinvest in structuring data for uses they can't predict yet, or underinvest and lose the ability to answer questions later. This removes that tradeoff. You store the data once, and you get to keep asking it better questions for as long as the models keep improving.

Getting started with Wasabi MCP

Performance Hub built MCP access into its own platform because it believed its customers deserved direct access to their own data, not just the reports the platform decided to generate for them. "We believe that our customers should be able to interact with the Performance Hub data via MCP," West said. "And now with Wasabi's MCP, Wasabi is offering to us what we offer to our customers."

Wasabi MCP is available now in beta, hosted or self-hosted, at no additional charge beyond standard storage costs. Connect an agent today, and find out what your own Wasabi data has been waiting to tell you.

Explore the quick start guide to connect your first agent. For a deeper technical reference, the Wasabi MCP documentation has you covered.

Not a current Wasabi account holder? Start your 30-day free trial or buy now.