What was the latest alarm on your machine? Which fault keeps coming back? What caused your last production stop?
To answer questions like these in your plant, you would usually have to go through historical data, find the relevant variables and analyze the results. Depending on the question, that can mean switching between several screens or tools.
With Savvy MCP, there is another way to work with your industrial data: you can ask questions in natural language and get answers based on the information already available in our AIoT platform.
What is Savvy MCP and how does it work?
MCP stands for Model Context Protocol, an open standard that allows AI applications to interact with external tools and data sources.
At Savvy, we have developed an MCP service that makes the query capabilities of our M2C API available to compatible AI assistants, including ChatGPT, Codex, Gemini and Copilot. The assistant can discover what information is available, retrieve the data it needs through the relevant tools and turn the results into an answer you can understand.
Savvy MCP currently enables you to:
- Discover machines and assets available in the platform.
- Explore variable and indicator descriptors to understand what data is available for each machine.
- Retrieve historical operating data and use it to answer questions.
- Query alarm histories and analyze incidents recorded over a given period.
What does MCP add compared with a traditional integration?
Our platform already gives you access to industrial data through an API. But using that data in an AI application often means building a dedicated integration.
MCP provides a standard way for compatible AI applications to discover and use the tools exposed by Savvy. This gives manufacturers, developers and system integrators more flexibility to build natural-language queries into their own applications, using the industrial data infrastructure they already have in place.
Three examples from a demo machine
To put Savvy MCP to the test, we connected an AI assistant to a demo asset called Machine SA. We then asked three questions without first explaining the underlying data structure to the assistant.
1. Find the latest alarm and the most frequent one
When asked about alarms, the assistant searched the historical records, identified an incident and retrieved the time of its latest activation. It also reviewed the previous 24 hours to find out which alarm had occurred most often.
Demo example: alarm history and activation frequency for Machine SA.
2. Discover which indicators are available for a machine
The assistant used the variable descriptors to identify the indicators associated with Machine SA. It found 46 active indicators and organized them into groups, giving the user a clear view of the data available for further analysis. This particular query returned the indicator catalog, rather than measured values.
Demo example: a grouped catalog of the 46 active indicators available on Machine SA.
3. Identify a shift pattern from historical records
Some questions require the assistant to interpret a sequence of data rather than simply retrieve a value. By reviewing shift changes recorded over several days, it inferred that the second shift ran from 14:00 to 22:00.
The assistant also made clear that this schedule was inferred from historical records, rather than read directly from a shift configuration. That distinction matters when using AI to interpret industrial data.
Demo example: identifying the second shift schedule from recorded shift changes.
Controlled access and usage monitoring
When AI assistants interact with industrial systems, you need to retain control over where they can access information.
That is why Savvy MCP lets you restrict access by tenant: a separate logical space within a Savvy Industrial Cloud or Fog deployment. This allows you to manage which tenants can use Savvy MCP while keeping their access separate.
The platform also provides usage statistics by user and MCP tool, together with additional consumption metrics, so you can monitor how the service is being used.
What's next: specialized MCP tools and connected workflows
We are working on expanding Savvy MCP's capabilities by developing specialized tools for specific use cases in production, maintenance, and quality.
These tools will perform targeted operations and analyses on industrial data, providing AI assistants with the relevant information for each query. This will allow us to address more complex use cases, streamline data processing, and reduce token consumption.
Furthermore, the MCP architecture allows a single AI application to use tools from multiple MCP servers. This creates opportunities to combine Savvy's capabilities with other tools within the same workflow, expanding the range of applications we can develop while maintaining our platform's existing integration methods with other systems.
With Savvy MCP, we are extending what manufacturers and system integrators can do with the industrial data already available in our AIoT platform.
Interested in exploring what Savvy MCP could do for your operations? Get in touch.