MCP
Model Context Protocol. An open standard that lets AI models and agents connect to external tools and data in a consistent way.
What is MCP?
MCP, the Model Context Protocol, is an open standard that defines how an AI application connects to outside tools and data sources. Instead of building a custom integration for each system, a provider publishes an MCP server that describes the actions and data it offers, and any MCP-compatible agent can discover and use them within the permissions it holds. In advertising, MCP is one of the protocols being explored to let agents query inventory, check availability or submit deal requests in a consistent way. It is complementary to A2A, which focuses on agents communicating with other agents rather than with tools.
How MCP works
A service exposes its functions through an MCP server. An AI agent connects to the server, sees what it can do and calls those functions with the access it has been granted.
- A platform builds an MCP server that exposes selected functions, such as listing inventory or requesting a deal, and the data each one returns.
- An AI agent connects to the server as an MCP client and authenticates with the access it has been granted.
- The agent reads the list of available tools and their inputs, then calls the ones relevant to its task.
- The server performs each action under its own rules and returns structured results the agent can act on.
Why it matters
In advertising it gives agents a standard way to query inventory or submit a deal, instead of a separate integration for every system.
For adtech platforms
An MCP server lets a platform offer one standard interface to many AI agents instead of maintaining a separate integration for each.
For publishers
Exposing inventory through a controlled MCP server makes it reachable by agent-led buying while the publisher decides exactly what is visible.
For advertisers
Agents working for a buyer can use the same method to talk to many sellers, which reduces dependence on bespoke connections.
Practical example
A buyer’s agent connects to a seller’s MCP server and requests the list of sites and formats it is allowed to see.
Key considerations
- MCP standardises the connection, not the business rules; pricing, eligibility and approvals still have to be defined by the provider.
- Access control and logging matter, because an agent can act quickly and at volume once connected.
- Use of MCP in advertising is at an early stage, and conventions for advertising-specific tools are still forming.
More AI & Agentic Advertising terms
Frequently asked questions
What is the difference between MCP and an API?
An MCP server usually sits on top of existing APIs. What it adds is a standard way for AI agents to discover which functions exist and how to call them.
How is MCP different from A2A?
MCP connects an agent to tools and data, while A2A is designed for one agent to communicate and coordinate with another agent.
Does MCP replace OpenRTB?
No. OpenRTB governs real-time bid requests and responses; MCP is more suited to tasks around the auction, such as discovery and deal setup.