Model Context Protocol (MCP) is starting to appear in contact centre technology, promising to make it easier for AI applications to connect with existing business systems. But what exactly is MCP, and how could it help contact centres? We take a closer look.
Think about how many different systems your contact centre uses every day. You’ve probably got a CRM platform, a knowledge base, workforce management software, and several other applications that your agents regularly switch between.
Now imagine if your AI tools could access information across those systems without needing a completely different integration for each one. That’s the idea behind Model Context Protocol (MCP), and it’s already starting to attract attention from contact centre technology providers.
In short, Model Context Protocol, or MCP, is an open standard that allows AI applications to connect with external systems, retrieve information, and use tools. Introduced by Anthropic in November 2024, it provides a common way for AI applications to communicate with different business systems.
Traditionally, connecting an AI application to another platform has often required a dedicated integration. MCP aims to make this process more consistent by providing a standard interface.
For example, imagine your contact centre uses one system to manage customer records, another to track deliveries, and a separate knowledge base containing product information.
With the right MCP connections in place, an AI assistant could potentially retrieve information from all three when helping an agent deal with a customer enquiry.
It doesn’t mean every system automatically works together, but it could make connecting AI to different applications considerably more straightforward.
For practical suggestions on how to develop an AI-positive culture, read our article: How to Nurture an AI-Positive Culture in Your Contact Centre
The basic principle is fairly simple.
An AI application connects to an MCP server, which makes certain information or functions available from another system.
The AI can then access those capabilities, provided it has the necessary permissions.
There are three main components:
Let’s put that into a contact centre example:
MCP provides a standard way for the AI to access those capabilities, although the underlying systems still need to perform the actual work.
Many contact centres have invested heavily in AI, but these tools aren’t always connected to the systems containing the information they need.
An AI assistant might be perfectly capable of answering a question, for example, but that isn’t particularly helpful if it can’t access the customer’s account details or check the status of an existing case.
MCP could help address some of these limitations.
Here are four areas where it could make a difference:
We’ve all seen how frustrating it can be when agents have to jump between different systems just to answer a relatively straightforward question.
Perhaps they need to check a customer’s account history, look up a payment, and then find the relevant company policy.
With the right MCP connections in place, an AI assistant could retrieve information from these systems and bring it together for the agent.
This could reduce the time spent searching for information, particularly when dealing with enquiries that involve several different applications.
For advice on how to avoid creating silos, read our article: How to Use AI to Connect the Dots – Not Create More Silos
Chatbots and virtual assistants are already widely-used in customer service, but their usefulness depends heavily on the information they can access.
A chatbot might be able to explain a company’s delivery policy, for example, but what happens when a customer wants to know where their specific order is?
Without access to the order management system, it may struggle to provide a useful answer.
MCP could make it easier for AI applications to retrieve that information from connected systems, allowing them to respond to more specific customer enquiries.
This could be particularly useful for organizations looking to move beyond basic FAQ-style chatbots.
There’s also growing interest in agentic AI, where AI systems can take action rather than simply provide information.
For contact centres, this could include updating customer records, booking appointments, creating support tickets, or initiating refunds.
MCP could help by giving AI agents a consistent way to access the tools needed to perform these tasks.
For example, an AI agent could potentially check a customer’s existing appointment, retrieve available alternatives, and submit a request to make a change.
Whether it could complete that process independently would depend on the permissions and controls in place.
Just because an AI application can access a system doesn’t mean it should be allowed to make every change without human approval.
MCP could also have applications beyond customer-facing AI.
For example, WFM teams regularly work with large amounts of data covering forecasts, schedules, staffing requirements, and operational performance.
With the right MCP connections, managers could access and analyse this information through AI assistants, without having to navigate multiple reports.
We’ve already seen examples of this emerging, with workforce management providers Assembled and Peopleware both introducing MCP capabilities that allow AI applications to access WFM data.
This could help managers analyse performance, explore scheduling improvements, and make better use of their workforce information, without constantly switching between systems.
For examples of where planning teams can make the most of AI, read our article: 8 Ways AI Is Already Changing WFM for the Better
No. MCP and APIs serve different purposes, although they can work together.
An API allows applications to exchange information or perform specific operations.
MCP provides a standard way for AI applications to discover and use tools and information made available by other systems.
In fact, an MCP server will often rely on existing APIs to retrieve information or perform actions.
The potential benefit is that developers may not need to create a completely different AI-specific integration every time they want an application to access the same capabilities.
However, there will still be development work involved, particularly when dealing with older systems or applications that don’t support MCP.
While the potential benefits are interesting, there are some important considerations for contact centres.
The first is security. Customer contact operations regularly handle sensitive personal and financial information, so organizations need to be careful about which systems their AI applications can access.
For example, allowing an AI assistant to check an order status is very different from giving it permission to change payment details or issue refunds.
There also needs to be a clear understanding of who is authorising those actions, how permissions are managed, and what happens if something goes wrong.
Another consideration is prompt injection, where malicious instructions hidden within information accessed by an AI application could influence its behaviour.
Organizations will need appropriate safeguards, including access controls, monitoring, and approval processes for higher-risk actions.
Then there’s the issue of data quality.
If your CRM contains outdated customer information, MCP isn’t going to fix that.
It may make the information easier for AI applications to access, but the underlying data still needs to be accurate and reliable.
Connecting an AI application to more systems doesn’t guarantee that it will make the right decisions or provide better answers.
MCP is still a relatively new technology, and contact centres shouldn’t feel they need to introduce it simply because more technology providers are talking about it.
However, it’s certainly something worth understanding, particularly if you’re exploring how AI could work more closely with your existing systems.
A sensible starting point would be to look at where your AI tools currently struggle to access information.
These are the sorts of problems where MCP might have something to offer.
It’s also worth asking your technology providers whether they support MCP, what information their integrations can access, and what security controls are available.
As more organizations explore agentic AI, the ability to connect AI applications with existing business systems is likely to become increasingly important.
MCP could make some of those connections easier to build and maintain. But for contact centres, the value will come down to what it actually helps them achieve.
If you are looking for more information on contact centre technology, read these articles next:
Reviewed by: Jo Robinson