Pagos expands its MCP server so merchants can query transaction data, approval rates, and fee breakdowns through Claude, ChatGPT, or Gemini. It is the first payments company to ship AI-native data access at scale.

Payments teams still export CSVs. They open spreadsheets. They filter, sort, and dig through raw transaction logs hunting for patterns that might be there. They send Slack messages asking analysts for answers. And they wait.

Pagos just made that entire workflow obsolete.

The AI-powered payments intelligence company announced a major expansion of its Model Context Protocol (MCP) server this week. Merchants can now access verified and harmonised payments data directly through Claude, ChatGPT, or Gemini using natural language queries. No SQL. No exports. No waiting for the data team.

The expanded Pagos MCP server ships with two core tools that merchants can plug into their existing AI workflows.

The first is BIN Database queries. Merchants can ask Claude or ChatGPT about card issuing banks, card types, brands, payment methods, and routing options. The system enriches live transaction data with verified reference information from card networks.

The second is Pagos Data queries. This is where the real power sits. Merchants can query their own transaction events, costs, and conversion data through plain English. Approval rates by processor. Fee breakdowns by card type. Processor performance comparisons. Decline code analysis. Pagos normalises field names like issuer descriptions and decline reason codes across different processors, so one query returns consistent answers regardless of which processor handled the transaction.

The implementation ships with no-code integrations and a Data Ingestion API for processor connections. Merchants connect their payment processor accounts, Pagos harmonises the data, and the MCP server makes it queryable. The remote server is live at mcp.pagos.ai/mcp.

Critically, data remains private and siloed by customer. One merchant's transaction history never touches another's. An architectural choice that matters in an industry built on competitive advantage and compliance requirements.

Pagos did not start from zero here. The company's AI chatbot has been a core platform feature for nearly two years. That domain expertise, knowing which queries matter to payments operators and how to structure data for reliable answers, is what separates this from a generic MCP wrapper.

Three shifts have collided to make this significant.

First, payments teams are getting leaner. Companies like Adobe, Eventbrite, GoFundMe, StubHub, Ultra Mobile, and Warner Bros. Discovery (Max) have all adopted Pagos to handle more payments complexity with fewer people. When your data team is not growing, you need tools that amplify what each person can do.

Second, business velocity has changed. "The days of flying blind without payments data visibility are done," said CEO Klas Bäck. "Payment operators want answers now, and they want them in their existing AI workflows." This is not nostalgia for the old way of working. It is pragmatism about competitive speed.

Third, AI is becoming the default interface for data access. Business intelligence used to mean dashboards. Then it meant SQL queries. Now it means asking an agent in the tools teams already use every day. CPO Albert Drouart put this plainly: "Payments teams are done spending hours digging through raw payments data. They need instant and actionable insights based on verified and reliable data."

The shift is not from dashboards to AI agents as an option. It is from AI agents as a nice-to-have to AI agents as the baseline expectation.

The Model Context Protocol is not new. Anthropic introduced it in November 2024 as an open standard for AI agents to connect to external tools and data sources. We covered MCP extensively in our MCP course for product leaders. But MCP adoption in fintech has been conspicuously slow. Most payments companies have watched the protocol from the sidelines.

Pagos is moving faster. By shipping an MCP server now, the company positions itself at the centre of what we have called the agentic commerce stack. This is not about building another SaaS dashboard. It is about recognising that AI agents need data the same way software developers need APIs.

The timing reflects broader momentum. On the same week, travel giant Agoda open-sourced APIAgent, a tool that converts any REST or GraphQL API into an MCP server with zero code. Visa and Mastercard are building AI agents that can execute transactions at machine speed. The entire commerce chain is shifting toward systems where buying decisions happen faster than humans can supervise them.

We covered five protocols racing to own AI commerce in an earlier analysis. MCP is one of them. Pagos shipping an MCP server demonstrates that the winners in payments infrastructure will be companies that position themselves as connectors in an agentic system, not just processors of transactions.

The real impact will not be analytics. It will be agentic workflows.

Today, merchants use the Pagos MCP server to ask questions. Tomorrow, they will use it to automate decisions. An AI agent could monitor approval rates in real time, detect when one processor is underperforming, and automatically route traffic to a better alternative. Another could analyse fee breakdowns by card type, identify categories where the merchant is overpaying, and surface negotiation opportunities. A third could correlate decline codes with transaction patterns and flag anomalies for compliance review.

These are not speculative use cases. They are the workflows payments operators already build manually. The shift from "ask the data" to "let the data act" is the next phase of the agentic commerce stack.

For payments operators, this creates a decision point. Build bespoke integrations with each AI framework, each processor, and each data source. Or standardise on MCP and let your agents connect through a common protocol. Pagos is betting the latter wins. And for the first time, merchants can test that bet without waiting for industry consensus.

When payments data becomes conversational, who controls the conversation: the merchant, the processor, or the AI agent asking the question?