Credit Card Companies Bank On Developers For Agentic Acceptance
Bangkok, Thailand – Jun 23, 2015 : Group of credit cards on computer keyboard with VISA and MasterCard brand logos
getty
Let’s go shopping. It’s a core human behavioral trait that many people find enjoyable and rewarding, while some even find it compulsive and addictive. But what if some of the shopping legwork was taken care of for us? Would that enable us to save time searching for the things that we want, would that allow us to avoid looking in locations that simply don’t have the products ot services we need… and would that make the “traditional shopping” activities that we carry out even more enjoyable as we concentrate on the endorphin-only purchasing that really makes us happy?
The credit card giants think so.
Multinational payment card services company Visa launched its Visa Intelligent Commerce in April this year to enable AI agents to browse and (where the user allows authority to designated AI agents) even purchase on behalf of humans. With a centralized view on how software application development professionals will integrate its new services into the global realm of e-commerce, the company was up front about the need to establish trust across users, organizations and banks with an appropriate level of protection against fraud and disputes.
In terms of next steps, the company is making a clear play for the hearts and minds of developers by this month offering a model context protocol server for Visa Intelligent Commerce. As previously detailed here, model context protocol or MCP is a standard protocol connection technology that enables AI systems to access and interpret the right context by linking them with the correct services, tools and data sources.
Developers can now connect to Visa Intelligent Commerce APIs through the organization’s MCP Server, so the statdardization aspect is there. Attempting to forge even deeper connections, the card behemoth is now piloting the Visa Acceptance Agent Toolkit and this is a technology built on top of the MCP Server. This no-code toolkit helps developers (and non-technical business users too, to a degree depending on the deployment use case) to work with AI agents using plain language. Visa insists that in this part of its total software tool offering, no coding is required.
Visa also announced new “stablecoin” partnerships (a technology amalgam of a fiat currency and a crypto that is pegged to a real world asset or commodity such as gold) across more geographies at the same time as expanding its Flex Credential platform, the latter being a set of products and services that provide ways for people to pay and get paid.
Trained Professionals, Please
What is perhaps most telling here – especially in the age of AI when so much of the discussion gravitates around who is going to automate themselves out of a job first by embracing artificial intelligence – is that Visa is putting developers first. The company appears to realize that even in the current climate of no-code drag-and-drop software tool democratization, this is a base-level technology that needs to be handled by trained professionals with a set of galavnized heat-resistant oven gloves.
“For AI agents and large language models LLMs to interact with Visa’s network, they need a secure, consistent way to communicate with our services. The Visa MCP Server makes this possible, starting with Visa Intelligent Commerce application programming interfaces. This is a major unlock for the developer community, as it provides a ready‑made integration layer for AI agents to securely access Visa’s capabilities with no requirement to hand‑code every API call or manage complex authentication flows. It enables’ developers to move from idea to functional prototype in hours,” stated Visa.
A merchant support software bot can respond to queries such as: “Create an invoice for $100 for [name of purchaser] that should have a payment due of the first of next month.” The AI agent uses the Visa Acceptance Agent Toolkit to call the invoice API, fill in all the required details and return a secure payment link with zero manual intervention or other development tasks required.
Global head of growth at Visa Rubail Birwadker says that developers want a faster path from idea to a secure, working agent. “With the MCP Server, AI agents can understand and call Visa APIs without hand‑coding auth or glue code, so teams move from weeks to hours. With our Acceptance Agent Toolkit, developers can prototype real agentic commerce workflows like invoicing with a prompt, then graduate to production with the same policy and permissioning model. It’s the shortest line from LLM to live transaction – built for builders,” he said.
AI Purchasing (+Marketing) Starts With Code
For those who have devoted their career path to programming, it may be encouraging to learn that Visa’s first touchpoint for agentic AI purchasing power starts with a self-hosted package for JavaScript developers. Analyst house Gartner studied this subject this year to build its report entitled “Build AI Agent Services to Revolutionize Client Operations,” a study that suggests organizations can achieve scalable flexibility using intelligent multi-agent systems and position an organization for competitive advantage with AI agents.
Mastercard is following a similar development trajectory with its Mastercard Agent Pay offering. Billed (pun not intended) as a means to “enhance generative AI conversations” by integrating trusted payment services, the organization is working with Microsoft to integrate Microsoft Azure OpenAI Service (enterprise-ready AI models with all the scalability groundwork also catered for) and Microsoft Copilot Studio (a graphical development environment used to build agents with sophisticated dialog creation, knowledge, actions and analytics) into its resultant platform of services.
Similarly focused on safety, Mastercard’s chief product office Jorn Lambert is on the record saying, “The launch of Mastercard Agent Pay marks our initial steps in redefining commerce in the AI era, including new merchant interfaces to distinguish trusted agents from bad actors using agentic technology.”
Across its developer portal pages, Mastercard details its approach to many of the core application development requirements that all programmers face, but with a specifically aligned view skewed to the card payments business.
“From manual testing to automated functional tests – and now AI-driven quality checks – the industry is evolving rapidly. But one area that often goes unnoticed is visual validation, which ensures that UI elements render correctly and remain pixel-perfect across different devices, browsers, and resolutions,” blogged Mastercard lead software engineer Harit Narke and Kedar Mohanty, the company’s worldwide director of software engineering.
Other Cards In The Pack
With the big two leading the charge, the et al field of American Express and Discover are of course joined by PayPal and CapitalOne as potential protagonists in this market; the primary cohort from China being made of Alipay, WeChat Pay and UnionPay. While the rise of agentic “buying” services will no doubt come from most or all of these organizations, their collective application of AI at this stage appears to be more focused on streamlining internal processes and fraud detection.
The corollary of this whole discussion must logically then be… online product and services marketing needs to start with agentic AI that markets successfully to agents rather than humans in the first instance. Financial Times reporters Cristina Criddle and Daniel Thomas in London wrote this August 2025 to highlight the work carried out by OpenAI, Perplexity, Google and Microsoft in this realm.
“[We now see] AI-powered features that allow users to search for products through chatbots, with autonomous agents able to complete orders on behalf of consumers. The rise of AI-powered agents has prompted sellers and brands to rethink how they sell products online, in particular how their products are spotted by AI systems and recommended by chatbots,” write Croddle and Thomas.
In terms of even wider views, Christopher Miller, emerging payments analyst for Javelin Advisory & Strategy has strong opinions on this subject.
“As the payments industry knows well, a network of interconnected parties is more powerful than a series of 1:1 connections… and the emergence of MCP offers an early opportunity to create such a network. Innovation is built on a series of small and large steps, so Visa’s launch of an MCP server is an important step forward. Launches like these are necessary advances that will help bring the broader agentic ecosystem to life, enabling agents to access networks and we expect to see other participants follow suit in the coming 6-12 months,” said Miller.
Shopping, Remember Shopping?
The bottom line and where the buck really stops in this discussion is secure usability for shoppers (remember shopping, back where this started?) and that will only come if we foster the development of a diverse ecosystem of specialized agents that are capable of “getting better” at how the shop over time. Starting with key developer cornerstones like plug-ins, use of MCP and interconnectivity through APIs, generalized agent shopping can then evolve into specialized vertical use cases.
Credit where it’s due then (pun intended this time), unlike cloud computing’s “let’s just turn it on and figure out security, suitability and sustainability afterwards” approach, the industry appears to be starting somewhere near the correct balance.