What Model Context Protocol (MCP) Means for Media Buyers

Mike Mullin, Senior Product Manager
AI has introduced a new three-letter acronym to the ad tech lexicon: MCP. It stands for Model Context Protocol, and if you've been anywhere near an ad tech conversation lately, you've probably heard the term. Mike Mullin, product manager at Index Exchange, breaks down how MCP fits into the bigger agentic AI picture and what it means for media buyers.

Get notified when new videos are available

What is MCP?

Model Context Protocol (MCP) is a standard that Anthropic developed and open-sourced in late 2024. At its core, MCP is a standard way for AI systems to connect to external tools and data sources and interact with them in real time.

Think of it as the next evolution of an API—a standardized way to ensure that different applications can understand each other when accessing tools and data.

Every AI system used to need its own custom integration for every tool or data source it wanted to reach. Now, with MCP, developers can build to one shared protocol instead.

Another helpful perspective is that MCP is like a USB-C for AI. In the same way that cable lets you plug all kinds of devices into all kinds of ports, MCP lets AI applications plug into external tools and data through one standard connection.

MCP doesn’t replace or rebuild the systems you already use. Rather, it creates a standardized way for AI agents to interact with them. Your existing platforms become more usable without needing to be rebuilt from scratch.

One thing to note is that you’ll often hear MCP and MCP server used interchangeably. MCP is the protocol itself—the shared language that defines how an AI system and a tool or data source should talk to each other.

An MCP server is what a company or platform builds using that protocol to allow access to its own tools and data. So when you hear that a platform has an MCP server, that means they’ve built an interface using the MCP standard. Every MCP server speaks the same underlying protocol, which is what makes this scalable.

Where MCP fits in agentic AI

Where does MCP fit into everything else you’re hearing about agentic AI? It helps to break down the ecosystem into three parts.

  1. MCP server: This is the plumbing, creating a standard way for an AI system to call tools and access data. Setting one up is becoming necessary in today’s AI world, but on its own, it’s not intelligence.
  2. Agent: An agent is the worker. It analyzes the data that an MCP server provides, diagnoses problems, and takes action, using judgment about when it needs your approval, and what to do next.
  3. Skill: A skill is one job an agent knows how to do, for example, troubleshooting a deal, building a report, or packaging inventory from a brief. Skills can be shared across agents, which is how this kind of intelligence scales.

MCP’s role in programmatic

MCP isn’t specific to advertising, but it has plenty of applications here. Programmatic has always relied on APIs to connect DSPs, SSPs, measurement platforms, and more, translating and sharing data across all of them. MCP builds on that same foundation, making it easier for AI systems to plug into the tools you already use.

That’s starting to change how buyers and sellers work. Instead of navigating multiple dashboards, you can query in natural language and get real answers, whether you are:

  • Creating audiences
  • Identifying and packaging inventory
  • Managing and troubleshooting deals
  • Setting up campaigns
  • Managing campaign pacing and delivery

Deal troubleshooting with MCP

Let’s look at one of those workflows in action: deal troubleshooting.

Say a deal stops delivering. Instead of digging through dashboards and logs yourself, you can simply ask your AI tool of choice—like ChatGPT, Claude, or others—”Why isn’t my deal delivering?”

Behind the scenes, the MCP server connects to the platform, retrieves the necessary data, and returns it to the assistant for diagnosis.

From there, an agent walks the bid funnel—checking deal status, DSP bidding activity, bid blocks, and creative audits—to find where things are breaking down, and surfaces the likely root cause.

But identifying the problem is only half of it. The agent also recommends next steps, like requesting a fix from the DSP seat, temporarily reinstating a previous creative to keep the deal delivering, or notifying the team that owns the asset.

You stay in control throughout. The agent surfaces the options and does the legwork, then you decide what actions to approve.

This is just one example, but it shows why MCP matters. An MCP server has quickly become table stakes. It’s the foundation that lets your tools and AI systems speak the same language.

MCP and the agentic future

Agentic advertising is in a state of rapid and constant evolution. The solid foundation provided by an MCP server will let you troubleshoot issues, discover opportunities, optimize campaigns, and act, all through the AI-powered interfaces you already use.

Standardization makes the agentic future possible, and getting the plumbing right is how we get there.

Learn how MCP and other standards are shaping the future of agentic advertising.