- Key insights: Klarna unveiled its Agentic Product Protocol, an open standard that allows merchants’ products to be discovered by AI agents.
- What’s at stake: Accurate product data is vital to ensuring that AI agents purchase what consumers intend.
- Forward look: Multiple protocols are emerging to guide agentic commerce.
As payment companies work to standardize
Processing Content
Klarna’s Agentic Product Protocol is an open-sourced protocol designed to complement Stripe and Open AI’s Agentic Commerce Protocol, but works with any protocol, including Google’s open sourced AP2.
APP’s core function is to make merchants’ products accessible by AI agents and interoperable, allowing those agents to know which products exist, where they are sold and other information such as price and availability.
For example, the protocol allows agents to search and compare the same product at multiple merchants, maintain product references over time for price alerts, and include other attributes like brand, model and condition in its reasoning.
“Klarna is investing in
“AI agents can autonomously guide product discovery, personalize recommendations, and streamline back-end processes, driving meaningful gains for both shoppers and merchants,” the spokesperson said.
The protocol is built on Klarna’s product graph and data infrastructure, which it says provides agents with instant access to more than 100 million products and 400 million prices.
Merchants can make their products available by either sending existing product feeds to Klarna’s hosted Agentic Product Protocol API or by implementing the open specification themselves and exposing their product data directly in protocol format.
“You do not need to be a Klarna merchant partner to integrate with APP, however, participating in the API requires a relationship with Klarna to specify how the data can be used and shared,” the spokesperson said. Merchants are required to maintain a consistent feed structure and regularly update product information.
The standard is a step in the right direction, but there’s still a lot of questions surrounding how merchants will want to interact with AI agents, said Aaron McPherson, principal at AFM Consulting.
“This looks like a promising standard, although in its current incarnation, it relies heavily on Klarna’s existing network of merchants, and of those only the ones that have opted in,” McPherson said.
Large retailers such as Amazon may want to keep their customers in-house within their own chatbots and may limit third-party agent access to their products.
If that were to happen, “a lot of product offers will be undiscoverable using this protocol,” McPherson said, but noted that it may allow smaller merchants to compete better with the big ones by leveling the playing field.”
AI agents disrupt merchant’s control of the customer experience, Sam Boro, a partner at Perkins Coie, told American Banker. Boro advises fintechs, banks, merchants and marketplaces on payment products and services.
“The merchant has dedicated lots of money and lots of time in building its strategy and its way of interacting with its customer, only to now have an agent read everything that the merchant has created and decide [how] to present the merchant’s products and services” to the customer, Boro said.
Protocols like this help standardize that information but also help to protect the merchant, he said.
“The benefit of these standardized protocols is that the protocol developers can create something that helps merchants to understand who they are dealing with,” Boro said.