What is the difference between MCP and A2A in advertising?

MCP (Model Context Protocol) connects an AI agent to its data sources and tools; A2A (Agent-to-Agent) connects one AI agent to another AI agent. In an advertising context, MCP is how a buying agent reads campaign briefs and brand safety lists; A2A is how that same buying agent talks to a publisher's selling agent to negotiate a deal.

MCP (Model Context Protocol): Anthropic's 2024 open standard for connecting AI models to external tools, data sources, and services; it defines how a model retrieves context and takes actions within its own system environment.

A2A (Agent-to-Agent): Google's 2025 specification for direct communication between autonomous agents from different organisations; it defines how agents discover each other, exchange messages, and coordinate tasks across system boundaries.

AAMP: The IAB Tech Lab's Agentic Advertising Management Protocols, which sit on top of A2A to define advertising-specific deal term structures, governance requirements, and bilateral record standards.

Protocol stack:The layered set of standards that together enable a buying agent to access its own data (MCP), negotiate with a publisher agent (A2A), and govern the resulting deal (AAMP).

What MCP does

MCP defines how an AI model connects to external resources within its own system. A buying agent using MCP can read live campaign briefs from a planning platform, retrieve brand safety lists from a verification provider, pull budget data from a finance system, and write deal confirmations to a campaign management tool.

Without MCP, an AI agent would need bespoke integrations for every data source it needs to access. MCP standardises these connections so that any MCP-compatible tool can be plugged into any MCP-compatible agent. It is the infrastructure for an agent to know things and do things within its own operator's environment.

In advertising, MCP is the protocol that keeps a buying agent informed: it knows the current brief, the available budget, the brand safety constraints, and the performance data from previous deals.

What A2A does

A2A defines how agents from different organisations communicate with each other. Unlike MCP, which is about an agent's relationship to its own tools, A2A is about relationships between independent agents operated by different parties.

A publisher's selling agent and a buyer's buying agent need to be able to find each other, authenticate their identities, exchange deal proposals, and confirm commitments. A2A provides the communication layer for this. It handles agent discovery (how does the buying agent find the publisher's agent), message exchange (how are proposals and responses structured), and session management (how is the negotiation tracked as a sequence of exchanges).

A2A makes it possible for a buying agent deployed by an agency to negotiate directly with a selling agent deployed by a broadcaster, without either party needing a shared platform as intermediary.

Where AAMP fits

AAMP provides the advertising-specific content layer on top of A2A's communication infrastructure. A2A defines how messages are sent and received; AAMP defines what the messages contain when the purpose is buying and selling advertising inventory.

AAMP specifies the deal term fields (inventory type, targeting parameters, pricing model, volume commitment, brand safety constraints), the negotiation sequence (enquiry, offer, counter-offer, commitment), and the governance requirements (bilateral deal record, audit trail, dispute resolution provisions).

An AAMP-compliant marketplace like Alkimi uses A2A for agent communication and AAMP for deal content, giving buyers and publishers a standardised way to negotiate and govern deals regardless of which agent platform each party uses.

Do buyers need to understand all three protocols?

Buyers do not need to implement these protocols directly. The agent platform handles protocol compliance. But understanding the stack helps buyers make sense of why agentic buying requires new infrastructure rather than just adding AI to existing programmatic tools.

MCP explains why an agent can read a brief without a human copying it across. A2A explains why an agent can talk directly to a publisher's agent without a human email exchange. AAMP explains why the deal that results from that conversation is governed by a bilateral record rather than a single-party deal ID.

Frequently asked questions

Can MCP and A2A be used without AAMP for advertising?

Yes, technically. MCP and A2A are general-purpose protocols; they can carry any content. AAMP provides the advertising-specific structure that makes the content of deal negotiations standardised and interoperable. Without AAMP, buyers and publishers would need to agree on custom deal term formats bilaterally.

Which protocol should advertisers ask their technology partners about?

Start with AAMP compatibility: does the platform support AAMP-structured deal negotiation and bilateral deal records? A2A support is a prerequisite for AAMP, so a platform that is AAMP-compliant will also be A2A-capable.

Is MCP only from Anthropic?

MCP was originally developed by Anthropic and open-sourced. It has since been adopted by a range of AI providers and tool developers as a common standard for tool access. It is not exclusive to Anthropic's models.