As companies continue to build towards an AI-driven future, agentic commerce has become a hot topic in the intersection of finance and technology. From online shopping and customer support, to digital banking, AI agents are being built to meet the demands of an increasingly AI-proficient world. These AI agents, powered by large language models (LLM), are capable of various tasks, including sourcing product comparisons, answering customer questions, or helping to choose the optimal flight to book. At Marqeta, we are inspired by how our customers are exploring ways to bring AI agents into their businesses to create powerful new experiences, and we’re excited to help make this vision a reality by enabling agentic payments through our technology.
Agentic Payments
For agentic commerce to truly reach its potential, payments must be responsibly incorporated into the workflows of AI agents while maintaining the existing high standards of compliance and protections consumers are afforded today. By enabling these payments workflows and data for AI agents, they can fulfill more complex tasks from start to finish. For example, they can go beyond simply surfacing the right flight options for a consumer, but also book the option the consumer chooses. Or an AI agent built to handle internal tasks can not only help surface details about a card transaction, it can also autonomously take the necessary action for managing a dispute.
AI agents must communicate directly with payment platforms and have the ability and autonomy to make it all work. And it’s not just simply making a payment. Agentic payments span from issuing a card to the rest of the user payment lifecycle, such as reviewing existing transactions, filing a dispute on a fraudulent transaction, and more.
Introducing Marqeta’s Model Context Protocol (MCP) Server
As part of bringing agentic payments to life, Marqeta’s Model Context Protocol (MCP) Server allows our customers to connect to our extensive payment APIs. An MCP server is an open-source standard developed by Anthropic for connecting AI applications to external systems, allowing them to access key information and perform complex tasks.
The MCP Server provides a standardized integration layer that enables AI agents to securely access and utilize Marqeta’s payment APIs. At launch, Marqeta customers can use the MCP Server to integrate payment workflows and accomplish the following tasks:
- Issue a virtual card: Instantly create and issue a virtual card that is ready for use at checkout
- Set spend and velocity controls: Limit the usage of the virtual card to specific parameters and give the agent the autonomy to be flexible and use its judgment where desired
- Dispute a transaction: Users can create and submit a dispute directly through the AI agent, pulling in Marqeta’s Disputes workflows to collect all of the required transaction data
- Manage and review transactions: Users can request more information about a transaction and ask the agent for help where needed
Example Scenario: Imagine a modern expense management company that has built an AI agent within their app. The agent handles a variety of tasks, including customer support, reporting, and managing new users. With a simple chat prompt from the user, the agent autonomously creates a virtual card for a new employee with precise spend controls, only allowing them to spend $50 per day and only at a select set of merchants. This card is instantly issued and follows the specific protocols set by the company without requiring any additional support or approval. Previously requiring manual input from the user, these actions are now all completed with an easy prompt.
And behind the scenes, this is simply done with an autonomous call from the AI agent at the time of the prompt to Marqeta’s MCP Server triggering the appropriate APIs to issue the card, set the spend limits, and anything else required.
Closing Thoughts
The future of agentic payments is full of possibilities and we look forward to supporting our customers in making sure Marqeta can be leveraged wherever they head next. For a more in-depth view of Marqeta’s MCP Server, check out our Marqeta Docs Page. We have also released the MCP Server as an open source project available on our GitHub repository, allowing developers interested in agentic commerce to experiment and explore what’s possible.
Are you ready to integrate agentic payments into your workflows today? Learn more about Marqeta's MCP Server.