WTF: Ad Context Protocol?

For some, it’s a bridge between today’s programmatic infrastructure and the dawn of the agentic era – in his Prebit Summit address earlier in the week, Scope3 CEO Brian O’Kelley compared its launch to the debut of header bidding – with comparisons to the OpenRTB standard. 

Here’s a breakdown of what AdCP is, how it works, and what it could mean for ad buyers, sellers, and everyone in between.

Okay, let’s get into the details 

At its simplest, AdCP is an open-source communication protocol that lets AI agents — whether built by advertisers, publishers, or ad tech intermediaries — interact using a common language.

Built on Anthropic’s Model Context Protocol, a.k.a. MCP, and other agent-to-agent, or A2A frameworks, AdCP standardizes how machines exchange structured data about audiences, inventory, and campaign objectives. In other words, it’s the digital plumbing that allows AI systems across platforms to “understand” one another.

Think of AdCP as the OpenRTB for the AI era. While OpenRTB standardized real-time bidding and ad exchange transactions, AdCP is designed to standardize agentic communication — the kind of back-and-forth negotiation that may precede a buy or audience activation, often occurring outside the traditional bidstream.

Why is it launching now?

The current programmatic ecosystem is full of complexity: data silos, opaque fee structures, and lengthy supply chains that obscure where budgets actually go. As companies begin developing AI agents to automate parts of this workflow — from planning to optimization — there’s a growing need for shared infrastructure that prevents fragmentation.

“AdCP gives the advertising industry a chance to establish shared technical infrastructure and standards that aim to accelerate progress,” said Adam Broitman, partner at McKinsey & Company, in the protocol’s launch statement.

The timing also aligns with the rise of agentic systems more broadly. Just as the OpenRTB standard helped unlock programmatic buying a decade ago, AdCP’s creators believe a machine-readable layer for AI agents could enable a new wave of interoperability — and perhaps even simplify the byzantine layers of ad tech that have built up since then.

How does AdCP fit with OpenRTB?

One of the first questions ad tech professionals will ask is whether this new protocol risks breaking existing systems. According to the consortium behind AdCP, it’s designed to complement, not replace, OpenRTB.

“AdCP is a parallel protocol to OpenRTB, intended to orchestrate advertising technology solutions, including DSPs and SSPs, that enhance their functionality through agentic interfaces,” said a spokesperson for the AdCP consortium. “Publishers and platforms can run both OpenRTB and AdCP simultaneously — they’re not mutually exclusive.”

In practice, AdCP enables buyer and seller agents to communicate criteria for a deal to happen, allowing for “direct” campaigns that can run inside a publisher’s ad server or through private and curated marketplaces.

That means platforms can adopt specific AdCP capabilities gradually, without having to tear out their existing stacks. “Think of it as adding a new lane to the highway, not tearing down the existing road,” per spokespeople advocating for the launch.

Built on widely adopted standards like MCP and A2A protocols, the design goal is interoperability, not fragmentation — a critical factor for an industry still grappling with multiple competing identity, privacy, and measurement frameworks.

Who’s behind it — and who keeps it neutral?

The founding members of AdCP include: Optable, PubMatic, Scope3, Swivel, Triton Digital, and Yahoo, with supporting members such as AccuWeather, Butler/Till, LG Ads, Raptive, Samba TV, and The Weather Company also on board.

But governance, not membership, may determine whether AdCP gains traction. The group says it’s launching as an open-source initiative, with a non-profit governing entity to follow soon.

“We’re actively forming a non-profit governing entity to ensure no single company controls the protocol’s evolution,” according to a spokesperson representing the collective. “The governance structure will include representation from publishers, advertisers, agencies, and ad tech platforms, with transparent decision-making processes.”

The roadmap and specification changes will follow an open contribution model, much like Prebid’s community-driven approach, with neutrality deemed foundational to this initiative. “This only works if everyone trusts the infrastructure isn’t tilted toward any one player’s commercial interests,” added the spokesperson.

That governance approach distinguishes AdCP from previous attempts to define new buying frameworks, many of which stalled when large vendors sought to protect their existing revenue streams.

Transparency and auditability?

The idea of AI agents negotiating and executing ad deals raises familiar worries: how can humans trust what’s happening under the hood?

AdCP’s designers say the protocol was built with auditability and bias control in mind.

“AdCP actually increases transparency compared to current programmatic systems by eliminating intermediary bid streams and obfuscating layers,” the consortium’s spokesperson explained. “Agentic transactions interact directly with sellers and their ad servers as true direct buys.”

Because AdCP is open source, any implementation must handle audit trails, data provenance, and identity interoperability as core requirements.

In one example, Swivel’s Seller Agent logs every conversational element with buyer agents, along with all objects created in the publisher’s ad server. The asynchronous design of AdCP — where responses can take seconds or days — also accommodates human-in-the-loop approvals, ensuring that automation doesn’t override editorial or brand safety judgment.

That design could address one of the biggest criticisms of programmatic systems: their opacity. If widely adopted, AdCP could create a layer of machine transparency — making the mechanics of agentic transactions observable and verifiable by both parties.

How will it change buying and selling?

In theory, AdCP could allow advertisers to brief agents in plain terms (“reach eco-conscious car buyers on connected TV in the U.S. this week”) and have those agents negotiate directly with publisher systems that understand the same schema.

Rather than auctioning individual impressions, agents could transact on audience segments, engagement rates, or even brand-lift outcomes. That opens the door to more flexible pricing and packaging models, while reducing manual workflow.

For publishers, it could simplify campaign execution by reducing dependency on intermediaries. Instead of selling through multiple exchanges and waterfall setups, publishers could expose their inventory and contextual data directly to agents — with transparent deal parameters and instant feedback loops.

What next?

The protocol is now available publicly on AdCP’s website, and the group behind it plans to expand its scope in 2026 to cover creative generation and performance attribution.

If it gains adoption, AdCP could become the connective tissue for the next wave of automation — much as OpenRTB did in the early 2010s. But for now, even its backers admit that success depends on broad buy-in from both supply and demand sides.

And as with any open protocol, neutrality and execution will be key. The history of ad tech is littered with well-intentioned standards that became dominated by a handful of powerful players.

Still, in an era where AI systems are increasingly negotiating, optimizing, and buying media on their own, AdCP could be the mechanism that lets those systems talk — and be held accountable for what they say.

The bottom line

AdCP represents an ambitious attempt to build the missing layer of AI interoperability in advertising. Its promise is clear — greater transparency, flexibility, and collaboration across a fragmented ecosystem. But realizing that vision will depend on whether the industry truly embraces an open, neutral infrastructure this time around.