MCP at One Year: Why Martech and Adtech Leaders Need to Now Pay Attention

By Dan Jaye, CTO

For years, “AI integration” has lived on roadmaps as a future initiative; something to prepare for eventually. That window is closing.

Model Context Protocol (MCP) just marked its first anniversary. In twelve months, it has evolved from an experimental concept into genuine infrastructure: multi-step workflows, secure authorization, composable extensions, developer SDKs, and a growing registry of servers across industries.

For those of us working in marketing and advertising technology, this shift matters – not in theory, but in architecture, cost structure, and competitive positioning. The time to engage is now. I’m excited about it.

MCP Is More Than a Connector. It’s an Operating Layer

MCP is often described as “a universal connector.” That undersells what it’s becoming: the operating layer between AI models and the tools, data, and systems that power the enterprise.

Everybody who is still at ‘I wrapped my legacy API’ is about three generations behind. The core reality is straightforward: custom wrappers and point-to-point integrations are accumulating technical debt. MCP is designed to eliminate them.

The anniversary release introduced several foundational capabilities:

• Task and lifecycle management for long-running workflows
• Dynamic client registration for secure authorization at scale
• Agentic tooling so servers don’t just return data, they perform work
• A formal extension system that allows additive functionality without bloating the core

This is hardly incremental improvement. It’s architectural maturation.

Three Breakthroughs Worth Understanding

1. Lifecycle Management for Complex Tasks

MCP now supports task states: working, required, completed, failed, and canceled, which allow AI to perform non-trivial, multi-step activities through tools. It’s not a traditional workflow engine, but it provides the state management that enables real execution.

It has a lifecycle, allowing non-trivial activities to be done by AI through tools.

This is the foundation for executable AI in marketing systems: segmentation, tagging, querying spend data, and even task orchestration across ad operations.

2. Security That Scales

You can’t pre-authorize every single client in the world. So dynamic authorization, or DCR, is critical because security has been maturing across the board.

Until now, AI agents and LLMs have operated without robust access control. MCP is the first standard advancing real security infrastructure – not aspirational frameworks, but working implementations.

3. Extensions Open the Door to Industry Standards

Extensions can be additive. They must not break the core. They’re modular and versioned independently. That opens the door to true MCP-based adtech standards.

This is the inflection point our industry should be watching. Extensions allow domain-specific intelligence without fragmenting the protocol. That’s how a marketing registry, campaign planner, or attribution layer becomes MCP-native.

Extensions are the Real Game-changer

Here are some critical extensions already available or in preview: 

1. Sampling with Tools: Agentic Servers!

This is the capability I have been anticipating since the June introduction of sampling and elicitation.  MCP servers can run their own server-side agentic loops, call tools in sampling requests, and allow concurrent tool execution.

2. Authorization and Secure Out-of-Band Interactions

These address a number of issues with authorization control at a fine-grained level (not every user should have access to every capability of every tool) as well as handling tasks such as securely obtaining user credentials during elicitation.

3. MCP Apps!

MCP UI was a game changer, allowing MCP Servers to become full featured applications, although restricted to iframe delivery.  Then OpenAI created their Apps SDK.  MCP Apps builds on MCP UI and OpenAI Apps SDK to deliver a standardized pattern for UI development within MCP and will be an essential capability for MCP deployments.

Why This Matters for Martech and Adtech Today

Our systems are already plural, fragmented, and overloaded. Most organizations operate ten to forty tools at once, with poorly defined integration boundaries, limited API depth, and no shared execution layer.

MCP offers the first credible path to coordination:

In other words, they’ve actually made it so servers can become agentic. It’s sampling with tools, where the server can do work. That distinction, from passive data retrieval to active execution, is worth sitting with.

The Registry Will Reshape the Competitive Landscape

The MCP registry isn’t just a directory. It’s becoming a new competitive layer. And we now need a sub-registry for advertising and marketing, because this is where marketing solution providers either step forward or quietly become commoditized. Once campaign tools, CDPs, identity graphs, tag managers, CRM sources, DSPs, and attribution engines register themselves, optimization becomes orchestration. That’s the turning point. 

The Cost of Waiting Has Changed

Our industry has experienced periods of stagnation before. I speak from experience: from 2002 to 2007, I focused my time outside of ad tech.  Then I came back, and nothing seemed to have changed. This moment will either repeat that pattern or break it. Let’s do the work.

MCP is ready enough. The problems it solves are real enough. The adoption curve has started. Our industry will either shape MCP’s next chapter or be shaped by it.

A Practical Path Forward

For martech and adtech leaders ready to engage, here’s where to start:

• Inventory candidates. Identify which tools, data sources, or processes would benefit most from direct AI execution.
• Stand up one MCP server. Even internally, connected to one real asset: a reporting database, campaign history, or identity graph.
• Define governance early. Establish credentials, permissions, and compliance boundaries before scaling.
• Watch the extensions space. This is where domain standards will emerge.
• Play offense. Don’t wait for vendors to dictate the path. Build once, reuse across every model.

The Ecosystem is Growing

The AI era won’t reward another generation of connectors. It will reward systems with shared context, unified protocol, agentic access, and secure execution.

That’s exactly what MCP is becoming. This time, we don’t have the excuse of unclear direction. The work is visible, the spec is public, and the ecosystem is growing.

The infrastructure is in place. The organizations that engage now won’t just anticipate the future of martech and adtech: they’ll build it.