Article Type: Release Notes Audience: All Users Module: Platform Releases
Note: QA Release Date: September 2, 2025. Production Release Date: September 19, 2025.
We're spending this month's release notes highlighting a new feature - support for MCP tools!
This month, we're starting the introduction of a new feature to enable agentic AI-driven applications in Fuuz - our own MCP server! This feature is currently in a closed beta state - we have some work to do before we're ready to roll it out for general availability, and some of the changes we're going to make will be disruptive, but if you want to request access to the closed beta to test the feature and provide feedback, please put in a ticket at support.fuuz.com and we'll review your request!
With this new service, Fuuz app developers and administrators will have the ability to use Data Flows to create custom MCP tools that interact with the Fuuz platform and apps running on it. Once the MCP tools are created, developers or end users can connect those tools to their preferred LLM chat agent, which will then be able to use those tools to do anything a data flow can do: retrieve, create, or modify data, initiate integration requests to external APIs, send notifications to Fuuz users, read data from shop floor devices, and more!
In this example, we've created tools that allow an AI agent to retrieve historical sensor measurements from shop floor devices. The agent is able to provide the appropriate filter parameters to the tools, then use the results to answer questions and provide insights about the data.

In this case, the MCP tool flow is very simple - it's essentially just a prebuilt query - but it could be more complicated if required. For example, it could retrieve live data from a device or external API, combine data from multiple sources, or pre-aggregate the results for easier consumption by the agent.

In the (sped-up) example below, we've created tools to return information about the schema of an app running in Fuuz, allowing the AI agent to explain the purpose of data models and their fields, their types, and how they relate. Some of the work we have planned involves integrating this kind of functionality directly into the MCP server, so it doesn't require data flows - along with extending it to the logical next step, allowing an AI agent to generate app components like data models or screens!

If I haven't driven the point home enough already, the takeaway here isn't "Fuuz can return sensor data or data models to an AI agent" - it's that the Fuuz platform has a new native feature for building custom agentic AI tools. Those tools could be an extension of an existing app - extending WMS to allow AI to assist with cycle counting, for example - or they could be entirely new MCP tools purpose-built to solve a problem in your business that's not easily addressable by existing AI tools, like driving insights from data stored in on-premise SQL databases or legacy cloud applications. We're putting the power in the hands of our users!
Naturally, with great power comes great responsibility - exposing features of an app to an AI agent comes with risk, and a lot of the work we have planned is aimed at guiding developers toward smart decisions regarding AI-driven functionality and providing administrators with the controls and transparency necessary to use it safely. The potential of these tools is incredible, and we're excited to get this feature into the hands of some of our forward-thinking customers and partners to see what they make with it!