For the agentic-curious: WTF is the Agentic RTB Framework?

WTF is Agentic RTB Framework

It’s a new technical standard for running real-time bidding that focuses on efficiency. It takes the existing RTB process and redeploys it inside a containerized environment so the different parts of a programmatic transaction can run much closer together. The setup lets any participant – no matter their role in te bid stream – introduce agents directly into the RTB system.

In plain terms please.

Think of it as a different way to run programmatic auctions. Now, most programmatic transactions involve multiple vendors and cloud services, each operating on separate infrastructure. Bid requests travel between those systems before a decision is made. Under the Agentic RTB Framework, that process happens locally within a shared computing environment, instead of across several external systems. 

Who’s behind the framework

A handful of major players have been shaping the framework over the last year including Index Exchange, Amazon Ads, Yahoo DSP, Basis, Chalice, Publicis Group’s Epsilon, Netflix, Optable, Paramount, WPP and others. Execs from Index Exchange have been particularly involved. Lending their  expertise in building containerized architectures and protocols that let agents operate within the auction. Chalice, meanwhile, was among the first to put that setup to use, developing an agent built to run directly inside the system the ad tech vendor created. 

Does that mean Index Exchange runs the show? 

No. The framework isn’t an Index Exchange product. It’s been shaped by what the company learned from its own development – what worked, what didn’t – whereas the version being built now is a broader, industry-led effort. 

So why does this matter for agentic? 

Because it finally creates the technical groundwork for agents – autonomous pieces of software – to plug directly into real-time bidding without the usual lag or risk. In today’s RTB setup those agents (whether they’re bidders, fraud checkers or data enhancers) have to communicate over the open internet, shuttling data through multiple systems and APIs. That setup makes real-time collaboration nearly impossible. 

“What we’re doing as an industry with this is effectively folding time,” said Anthony Katsur, CEO of IAB Tech Lab.

How does this happen? 

The framework uses containers – portable software units that hold things like bidding models, fraud checks or data processing logic. These containers can operate in the same data center, or even on the same server, as where the auction takes place. That setup allows data to move within a local environment rather than travelling across the open internet. Communication happens through faster, bi-directional protocols like gRPC, instead of traditional HTTP calls that open and close for every request. 

“We want you to take your black box – the magic that you do really well – and express a series of changes that you want to make into the transaction without you having to build all of that infrastructure independently,” said Joshua Prismon, chief architect at Index Exchange.

And the containers use the framework to talk with one another?

Correct. The framework is a standard way for these containers to communicate with one another during the bidding process.

Why is that important?

Running everything in one environment cuts latency, shrinking a transaction that used to take 400 to 600 milliseconds down to as little as a fraction of that. 

Why does lower latency matter? 

Faster transactions free up time to add real-time enrichment, identity data or fraud analysis before the impression is won – functions that normally wouldn’t fit in the current time window. 

Does it lower costs?

Yes. Because data says inside the same local environment it avoids cloud egress fees and reduces the number of open network connections. 

Is this a big deal for advertisers?

Remember those containers? They give advertisers a new level of control over how their bids are made. With this framework, they can wrap their own bidding logic – the models that decide how much to bid and when – inside a container that runs directly in the same environment as the auction. No more relying on the slower, restricted APIs that DSPs offer today. 

And? 

For advertisers, that means becoming true bidding agents. Their models can score impressions, devices and audiences in real time, using their own rules instead of a platform’s default. As Chalice CEO Adam Heimlich put it, this is the first time they can “bid exactly the way an advertiser asks”, without being held back by someone else’s infrastructure. 

What about publishers?

They stand to gain from the same efficiency. With decision and data processing happening in the same place as the auction, they can add more signals to each bid request in real time — things like contextual data, audience information or fraud checks — before it ever leaves their system. 

“That could include real-time fraud investigation, ” said Katsur. “Today, most fraud is dealt with after the bid but this approach could let us detect and address in real-time or at least take a more aggressive sample before the transaction is processed.”

In effect, the framework makes it possible to enhance the bid request — the message they send to solicit bids — not just send it downstream. In this scenario a publisher could work with their partners to enrich or verify data on the spot, making each impression more valuable and trustworthy. 

Speaking of collaborations, tell me more? 

Because sensitive data doesn’t have to leave the container, it opens up new ways for companies (banks, CTV sellers or data providers, for example) to collaborate securely without exposing raw data. 

How? 

Containers can be cryptographically signed to prevent tampering, and the framework includes permissioning and intent-based controls to ensure that agents behave as expected. 

“Its important to cover the potential for data that’s been on the sidelines that could come into play thanks to the framework,” said Heimlich. “That could be seller sof CTV who’ve been reluctant to send show level or episode level data or banks with purchase data. They can use containers to collaborate without losing control of that sensitive data.”

When will it launch? 

Not anytime soon. The framework is out for public comment, giving the industry until mid-January to weigh in. After that, the IAB Tech Lab will sort through the feedback, make revisions and start rolling out changes. The goal is to have the first official wrapped up by the end of the quarter but as always, that depends on how fast consensus comes together. 

Will it work with Ad Context Protocol (AdCP) and other standards?

Yes. The framework is built to work alongside existing protocols, not replace them. That includes AdCP and the User Context Protocol (UCP), which standardize things like data sharing, audience discovery and identity matching between agents.