Why AdCP

Why AdCP exists: the fragmentation problem across RTB, platform APIs, and direct IO — and how a universal agent protocol for advertising solves it.

Allocation, Not Day Trading

RTB treats advertising like day trading: “What is this impression worth?” It works for fungible inventory but commoditizes everything.

AdCP enables portfolio-level allocation: “How should I invest my ad budget?” This mirrors how advertisers actually think—they buy outcomes, not impressions.

Agentic Advertising is for Allocation

Why the RTB mental model doesn’t fit how advertisers actually make decisions.

The Structural Problem

Programmatic advertising defaults to a single question: “What is this impression worth right now?” That framing embeds CPM-based, auction-centric assumptions into the transaction layer — assumptions that do not hold for most global ad spend. Direct deals, sponsorships, broadcast, out-of-home, and retail media all operate on different pricing models, different timelines, and different success metrics. They are not edge cases; they are the majority.

AdCP sits at the campaign-workflow layer, above the auction. It does not replace OpenRTB or compete with serve-time decisioning — it coordinates the planning, negotiation, proposal, and deal-creation steps that happen before and after the auction fires. The protocol comparison explains how AdCP and OpenRTB coexist.

By structuring these workflow steps as protocol-level tasks, AdCP enables agents to apply brand identity, content standards, and governance checks at decision time — not as an afterthought. Agents that use these tasks can reason about brand suitability, creative fit, and compliance as part of the buying process rather than bolting them on post-execution. The protocol makes this possible; individual agents decide how much of it to use.

The Fragmentation Problem

Today, advertisers face three completely different buying systems:

Paradigm	Era	How It Works
RTB/Biddable	Legacy web	Real-time bidding via OpenRTB
API-based	Modern social, AI	Platform-specific APIs (Meta, TikTok)
Direct IO	Legacy	Insertion orders, manual deals

Each requires different integrations, tools, workflows, and expertise. 90% of ad spend never touches RTB—it lives in walled gardens, direct deals, and premium inventory.

AdCP creates a universal API standard that unifies all three under one umbrella.

Omnichannel By Design

Buying billboards is fundamentally different from buying social links:

• Different pricing: Flat rate vs CPM vs engagement-based
• Different creatives: Static images vs dynamic video vs conversational AI
• Different measurement: Impressions vs engagement vs footfall

AdCP creates a conceptual layer that abstracts these differences while preserving what makes each channel unique. One protocol, any channel.

Why Agents?

Intelligent agents reduce the cost of managing complex, negotiated deals:

• Adapt to nuances without over-specifying everything in code
• Handle variability across platforms and channels
• Natural language lets buyers describe intent, not configure parameters
• Scale relationships from 3-5 platforms to 20+ without scaling teams

The Protocol Layer for AI

AdCP isn’t just unifying legacy systems—it’s the protocol layer for emerging AI surfaces.

Sponsored Intelligence

Like VAST defined video ad serving, SI defines conversational brand experiences in AI assistants. When an AI says “Delta has flights to Boston—want me to connect you with their assistant?”—SI defines what happens next.

Sponsored Intelligence and the Trillion Dollar Sentence

How conversational AI changes the economics of advertising.

Brand identity

Standardized brand identity for AI-powered creative generation. Brands express who they are—colors, tone, assets—in formats AI systems can consume.

Together, these support fully AI-powered systems like Performance Max that need structured brand inputs, not manual campaign configuration.

Sponsored Intelligence

Monetizing AI surfaces — the reversed data flow, product spectrum, and SI Chat Protocol.

Brand identity

Standardized brand identity for AI creative generation.

Design Implications

These goals drive AdCP’s technical design:

• Asynchronous: Deals take time. This is not a real-time protocol—operations may take minutes to days.
• Human-in-the-loop: Some decisions need human approval. Publishers can require manual sign-off.
• Multiple transports: MCP and A2A provide the same tasks through different protocols.

The Protocol Family

Protocol	Purpose	Key Tasks
Media Buy	Campaign execution	get_products, create_media_buy
Signals	Audience targeting	get_signals, activate_signal
Creative	Ad creative management	build_creative, sync_creatives
Governance	Brand suitability, quality, compliance	Property lists, content standards
Sponsored Intelligence	Conversational brand experiences	si_initiate_session
Curation	Inventory packaging	Coming soon

Next Steps

Protocol Comparison

MCP vs A2A—when to use which, and what’s the same.

Security Model

Why agentic advertising raises the stakes, what AdCP defends against, and a checklist for brand IT and CISOs.

Client SDKs

JavaScript and Python libraries for AdCP.