A connective infrastructure called Model Context Protocol (MCP) allows large language models (LLMs) and other types of AI to interact with tools and data sources. This open-source framework facilitates these connections between tools.
“Model Context Protocol (MCP) is an open standard that enables AI systems to connect directly to the tools, data, and services where real work happens,” said Fred Roma, senior vice president of product and engineering at MongoDB.
AI models have been trapped inside chat interfaces and limited in their actions, according to Roma. Now MCP makes applications AI accessible for software builders. That enables agents to discover and use external tools, he added.
“Instead of building a custom integration for every app, builders expose capabilities through an MCP server once, and any compatible AI can find it and start using it,” Roma said.
You can connect a platform like ChatGPT to a common tool like Slack. And if you connect MCP to Atlassian Suite, Jira or Confluence, you can prompt AI to perform tasks with those systems, according to Steve Buchanan, engineering manager, cloud platform (Azure and AKS), at mobile device management company Jamf and a Pluralsight author.
“It’s like an API but for AI,” Buchanan said.
Experts noted how MCP is often described as the USB-C standard for AI.
“It creates a single shared interface across tools like MongoDB, GitHub, Slack, Google Calendar and beyond, so agents aren’t dependent on bespoke plumbing for every system they need to touch,” Roma said.
MCP addresses the problem of large language models not being able to interact organically with the outside world, according to Lyron Andrews, author fellow for Pluralsight and multidisciplinary service provider for Profabula, which offers IT management and professional development.
“MCP is a connector that enables the LLM to access up-to-date information,” he said.
Andrews explained that MCP allows users to connect to applications without writing new code. It also acts as the “hands and legs” for AI agents because the MCP directs the agent to operate in a certain area.
“When you tell that agent to go outside of the original scope (its original perimeter), you're going to use MCP to do that,” Andrews explained. “And MCP is going to do it in such a way that it can extend the agent from its internal world to go out and grab tools necessary to present relative contemporary and updated information.”
How Tech Professionals Apply MCP
MCP is useful in fields such as finance, healthcare, and IT operations to connect key tools. For example, health systems would use MCP servers to connect scheduling, lab results and insurance systems.
In healthcare, MCP servers would allow EHR vendors to connect AI systems, according to Buchanan. In addition, organizations can transport AI notes from customer service calls into CRMs, he said.
In IT, MCP servers allow DevOps and IT operations to come together. MCP servers could allow SRE agents to enter a conversation in a Jira ticket.
“[If] I want to capture all of that information in a Jira ticket, an MCP server in between Azure and my Jira system is going to be the glue to make that happen, to take that conversation and put it in Jira,” Buchanan explained.
In addition, MCP servers allow apps such as Zoom to connect voice recording and AI note-taking tools.
In addition, given that MCP servers determine what AI can do, designing these servers is an important part of implementation, according to Roma.
“The tools you expose, the permissions you allow, and the systems you connect all shape the scope of the AI’s behavior,” Roma said.
Companies such as MongoDB and Dice have MCP servers. Dice launched its MCP Server in January to connect AI assistants such as Claude, ChatGPT, and Gemini to its tech-only job database.
Similarly, the MongoDB MCP Server allows MCP-supported clients like Claude, GitHub, Copilot or Windsurf to interact with MongoDB deployments. It uses natural language and performs database operations with a user’s preferred agentic AI tools, assistants, and platforms, Roma said.
“What’s powerful here is the shift in how developers work,” Roma explained. “Instead of juggling multiple tools or building custom integrations, they can stay in their development environment and use intelligent LLMs to interact with live, context-rich data in real time.”
How MCP’s Open-Source Connectivity Is Important
With MCP being open source, it serves as a “foundational infrastructure” rather than a “proprietary integration layer,” Roma explained. Organizations can integrate multiple AI tools into their workflows without being locked into a single platform.
It also allows developers to study how the protocol works, contribute to it and build servers and clients that are compatible, he added.
“That openness gives organizations more visibility into how data is accessed and how actions are carried out, which matters for trust, compliance, and governance,” Roma said.
Buchanan noted that Anthropic developed MCP as an open-source protocol.
“Because [Anthropic] made it open source, now software vendors out there can develop their MCP servers on that MCP protocol,” Buchanan said.
With its open-source capabilities, MCP provides interoperability and universal access to platforms, according to Andrews.
“It allows you to be in a vendor-neutral state, which gives the consumer freedom of choice so [organizations] can go and access a variety of platforms, resources, and tools,” Andrews said. “It stops or disrupts this fragmented custom integration.”
How to Learn MCP
To learn MCP, tech professionals should jump right in and use it, Roma advised.
“Set up an MCP server, connect it to an MCP-compatible client, and see how a natural language prompt becomes a structured action,” Roma recommended. “Once you see that flow end to end, the protocol becomes much easier to understand.”
In addition, Roma suggested that tech pros should learn practical skills like backend development and AI design, as well as become fluent in the data systems that AI agents need to access.
Buchanan recommends using the tooling in Docker Desktop to use MCP servers.
“You can also start building AI agents and connecting those to MCP servers, because if you have AI agents, the MCP servers become the glue between different systems,” Buchanan says. “Docker Desktop has simplified all of that to make it easier, so you don't have to be a developer, like doing things under the hood [or] building your own MCP servers from scratch.”
Buchanan recommends that tech pros learn about development debugging and tools like Visual Studio Code. He also recommends learning Python to prepare for building MCP servers.
Andrews notes that Anthropic has a helpful sandbox environment for learning MCP.
To learn MCP, Andrews recommends taking a Pluralsight course on the path for learning MCP. The course helps bind theory into actual practice, he noted.
MCP holds potential to make it easier to have AI agents deliver specific data on a dashboard, just as a food delivery service like Uber Eats would deliver food, according to Andrews.
“It's a huge impact on efficiency, on speed, and on capability that you could bring with this universal connector of MCP,” Andrews said.