AdCP Is an Open Standard for Agentic Advertising

Triton Digital is one of seven companies that have launched a standard called Ad Context Protocol, or AdCP. The founders also include Yahoo, ebiquity, PubMatic, Scope3, Swivel and Optable.

“The advertising ecosystem is fragmented,” they state on the AdCP website. “Every platform has its own API, its own workflow, its own reporting format. Media buyers and agencies waste countless hours navigating this complexity.”

The companies hope the standard will unify advertising workflows across all platforms. “Think of it as the USB-C of advertising technology,” they state.

To learn more about it we asked Jean-Luc Wasmer. He is vice president, architecture and partnerships, at Triton Digital, which he joined in 2011. Since 2021 Triton has been part of iHeartMedia. This article is from a Radio World ebook.

Radio World: Before we dive into AdCP, how is programmatic buying of broadcast radio ads going? When we spoke more than two years ago, you had just made iHeartMedia OTA inventory available on the Triton Audio Marketplace. How widely is programmatic now being used?

Jean-Luc Wasmer: The OTA marketplace is used daily and has grown a lot in the past two years. When we launched, we had a single buying platform, or “DSP,” integrated in our exchange, or “SSP”. We now have six more DSPs, including Yahoo and Google’s DV360. This allows more buyers to take part in the auctions, which increases competition and with that, the price of the ads.

At the same time, we also worked with Katz Radio Group to onboard other U.S. broadcasters. 

Most of them already had Jelli appliances but they hadn’t been used in a while, so they had to be recertified on a case-by-case basis. This was more time-consuming than expected but we’re happy to announce final production tests are well underway and we’re planning to launch Katz-represented broadcasters in Q1 2026. This has been requested not only by broadcasters who wanted to be part of the marketplace, but it’s also something buyers have asked for.

RW: So, what prompted the AdCP initiative?

Wasmer: Triton and others in the industry realized that the ad-tech innovations of the past two decades — and especially the rise of programmatic — have increased operational complexity. Setting up campaigns involves a lot of repetitive, manual work: multiple screens, countless settings and configuration steps. We asked ourselves: How can we use AI to simplify all that? 

We believe we’re entering the era of “agentic advertising.” We’ll start seeing more AI agents assisting with advertising workflows, helping reduce the layers of complexity that have accumulated over the years.

But to make that possible, we first need a communication protocol that standardizes how AI agents interact with each other. And it can’t just be Triton and a handful of companies defining it — it needs to be something the entire industry can adopt. That’s why we created AdCP as an open standard, built on top of another open protocol, the Model Context Protocol.

MCP was released almost exactly a year ago to simplify how large language models connect to external systems. Recall that at first, LLMs could only answer questions based on their training data. Today, applications like ChatGPT can access live information online. Not only can they search the open web, but they can also query databases and interact with external tools. You can tell an LLM: “Find this data, then send an email, create a reminder,” and so on.

With AdCP, we extended MCP by adding a layer to define advertising-specific data and tools. Our hope is that others will adopt it and contribute to the project.

RW: Will agentic advertising replace programmatic?

Wasmer: No. Programmatic made the buying of ads dynamic by introducing real-time auctions. Of course, auctions are fully automated — they last about 100 milliseconds — but the rest of the workflow, what happens before and after the auction, is still manual. The goal of agentic advertising is to automate most of the tedious operational tasks involved in running ad campaigns. This includes finding the inventory, negotiating deals, setting up the targeting, assigning the creatives, monitoring the performance of the campaign and adjusting parameters as needed.

Take inventory discovery for example. Today, someone has to manually search across stations and podcasts to find the shows that match an advertiser’s audience. Buyers also need to understand that targeting a podcast requires different parameters than targeting a stream. This work is time-consuming — often days or weeks. It is resource-heavy and prone to human error. AI is going to excel at this.

With AI agents, a buyer can simply focus on the target audience and ask, “Where should I be advertising this product?” The buyer’s agent can communicate with the seller’s agent, which returns the appropriate inventory and audience details. The idea is to let people focus on their real job — whether it is marketing products (buyers) or selling ad space (broadcasters) — instead of wrestling with interfaces and data entry.

From an architectural point of view, AdCP lives within a new AI agent layer — in green in the accompanying image — that will sit above the current ad delivery tech, shown in orange.

And importantly, AdCP doesn’t dictate the buying method — it works with both programmatic and direct/IO-based transactions.

We now have a few demo systems in place and we’ve started to build this into our products. We believe the whole industry is going to shift fast because this will make things way more efficient. 

This article is from the ebook “Streaming Best Practices.” Access it here.