What is an MCP server? And how can medtech companies use one with their QMS?

It can be surprisingly difficult for businesses to get accurate answers to questions about their own data. Often, the data isn’t easily accessible to employees without specific technical skills or it may live in multiple different systems. Some data may be in Salesforce that isn’t in the company's content management system (and vice-versa), but an employee may need all of that data to get the full picture.

For many organizations, an MCP (Model Context Protocol) is the answer to this problem. Many employees are already using LLMs like Claude or ChatGPT to some extent throughout the day. An MCP simply increases what they can do with these applications, allowing them to get answers about internal data sources using the same method they’ve grown used to: asking AI.

BONUS RESOURCE: Click here to download your free QMSR readiness playbook, packed with resources to help you prepare.

Why would you need an MCP server?

If you were to ask an LLM like Claude to connect to your database and generate some insights into a certain data set, it will tell you it can’t do that because it doesn’t have access to that database. The data is outside its context. This will happen for a wide range of specific requests that require the LLM to get information from a tool that is outside its context, like calendar, mail, or document storage apps.

This isn’t just a minor inconvenience for individuals. Most businesses have huge amounts of data and files living in tools like Salesforce, Dropbox, or similar systems. And while their employees may regularly use these tools and use LLMs like ChatGPT or Claude, they typically aren’t connected.

By using an MCP server, organizations can eliminate these silos and allow AI applications to securely access the tools and databases where information is stored.

What does an MCP server do?

The MCP acts as a bridge between the AI application and internal data sources you want to use. For instance, let’s say you wanted to understand the frequency of an activity captured in your internal database, or look for patterns in your data. You could ask an LLM a question about your data, and it would use the MCP to securely access the database and bring you back an answer. This is especially powerful because the AI can then format the answer in a manner that’s easy to read and digest for humans (rather than keeping it in lines of code).

Essentially, you can think of the MCP as a USB port for internal data. An AI application like ChatGPT can plug in to data sources, tools, or workflows and use that access to translate your internal data into a structured format.

How does the MCP work?

The MCP requires three basic players:

• The client: This is the primary application where the user interacts with the AI, typically an LLM like Claude or ChatGPT.

   

• The server: The MCP server securely exposes specific tools (functions the AI can perform) and resources (data the AI can read) to the client.

• The external tools: These could be any number of different tools that are connected to the host via the MCP. They could include an internal database, a CRM, a calendar app, a booking app, etc.

Essentially the client (the LLM), communicates to the server through the MCP, telling it what to retrieve from the connected tools. It’s known as an MCP server because the model (the AI application) gets additional context from a tool via the protocol, which is simply a standardized means of communication between the client and the server.

What are some common uses for an MCP server?

MCP servers are growing in popularity because they are enormously useful for getting accurate answers to specific questions about large amounts of internal data. Instead of attempting to search the database themselves, users can now interact with an LLM in the manner they’re used to (simply asking what they want to know) and will receive an answer that is based solely on the internal data the MCP accesses.

And because MCPs can act as a connector to multiple different data sources, they can help eliminate much of the context switching that employees have to do if they are moving from one system to another to find answers. If the AI is connected to, for example, Slack, Jira, and the company’s CRM via the MCP, it can pull from all of these sources to answer questions posed by a user.

How could you use an MCP server with QMS software?

Here’s where it gets interesting for medtech. Most medical device companies don’t just use Slack and Google Drive. They have a huge amount of internal data they are legally mandated to collect and manage in the form of their quality management system (QMS).

If a company were using an eQMS solution, that tool could be connected to an LLM via an MCP server. Theoretically, you could then ask a question in your LLM like, “What training has been completed in the past 30 days?” Using the MCP server, the LLM could get the answer from the eQMS and, crucially, format the data in a way that’s easily understandable to humans, such as a table. You might then have the LLM create charts or decks to make management review prep easier, or draft emails to everyone who needs to complete a training. The same goes for questions about CAPAs, suppliers, nonconformances, and anything else you’re documenting in the eQMS.

The benefit of using an MCP server in this situation is that your QMS becomes more than just searchable. You can ask it to bring you different data for comparison, help identify patterns, and transform the way you think about your quality system. And you’d be able to do it all using the same AI tools you’ve grown used to using over the past few years.

BONUS RESOURCE: Click here to download your free QMSR readiness playbook, packed with resources to help you prepare.

Greenlight Guru's modern software solutions are built specifically for medtech

The pace of innovation around AI is moving fast. And while it’s important to know you’re getting the latest features and functionality in the software you use for medical device development, quality management, and clinical trials, it’s also important to know those features are secure and effective for medtech.

That’s why at Greenlight Guru, we’re always looking at what makes sense for our customers, meaning what will help them build safe, high-quality medical devices that change patients’ lives. If you’re ready to see what the Greenlight Guru Platform can do for your medical device company, then get your free demo today!