With various consumer applications of artificial intelligence taking hold, it was only a matter of time before expectations for personalization and insight migrated to the banking world.
Narmi has launched a groundbreaking ability for end users to integrate with popular LLM models through a MCP (model-context protocol) server. This integration allows users to ask wide-ranging questions about account activity and receive suggestions meant to help them build savings, track spending, and empower their financial lives with data made available securely through their financial institution.
Adhering to industry standard compliance and risk practices, Narmi’s MCP server was built to address the needs of a growing segment of the market that wants to use artificial intelligence to help with their personal finances.
As the first product of its kind developed in the US, Narmi’s MCP integration will serve community financial institutions looking for a head start in driving primacy by closing the credibility gap for smaller banks and credit unions.
Narmi's MCP server enables users working with LLMs like Claude or Gemini to query their financial information directly. Users can ask questions like "What's my checking account balance?" or "How much did I spend on groceries last month?" and receive instant responses generated directly by their LLM.
Through securely connecting the power of LLMs with relevant financial context, consumer and business banking users now have access to an always-on personal financial manager, without the cost of one.
"We don't see AI as a trend – it’s innovation that can directly address the unique needs of community banks and credit unions. This integration represents the first time customers can analyze their financial data through an LLM like Claude, making banking as intuitive as asking a question. It's exactly the kind of innovation that helps our banking partners compete with the biggest names in finance."
— Chris Griffin, Co-President and Co-Founder, Narmi
Our MCP server enables the following services for customers at community financial institutions:
Users can now leverage AI to analyze their financial information, ask questions about their finances, and generate reports, ranging from monthly budgeting recommendations for consumers to detailed cash flow analyses for businesses.
For example, Ethan, a 29-year-old professional and customer of a community bank, is planning his monthly budget. Instead of logging into his financial planning app and manually categorizing expenses so he can understand the data, he opens Claude on his desktop and asks: "How much did I spend on restaurants and entertainment last month, and how does that compare to my income?"
Within seconds, Claude responds with his spending breakdown, shows the percentage of his income these categories represent, and suggests a realistic budget adjustment based on his actual transaction data. Ethan can then ask follow-up questions like "I need to cut my spending. What subscription services am I paying for?", creating a conversational financial planning session.
"What excites me most about this feature is how it democratizes access to sophisticated financial insights. Previously, only customers of some banks, or people who paid for financial planning apps, had access to AI-powered financial tools. Now, every community bank can offer their members the same cutting-edge experience, quickly leveling the playing field in a way we haven’t before."
— Yaro Melnyk, Group Technical Product Manager, Narmi
Narmi’s MCP server ensures seamless integration that maintains the bank's brand identity and custom deployment options, including api.mcp.[bankname].com domains.
Our implementation of the MCP server follows bank-grade security protocols to protect sensitive financial information, and users must authorize LLM access to their data with every new chat.
The setup process follows three key steps. The result is a powerful experience through Narmi’s MCP server that allows users to leverage the power of AI models to analyze their finances without storing any sensitive data and maintaining bank-grade security protocols. Below is a high-level overview of how a user can set this up with Claude Desktop:
Please note: This is a summary. For a complete step-by-step flow, including all necessary code and configurations, please see our Developer Docs.
The first step is to generate your API credentials that will allow Claude to connect to your banking system. Once you have followed the steps to create an API token, securely save both the API token and secret. Treat these credentials similarly to your digital banking user name and password - you will need them in the next step.
Next, download the client.py file and configure your Claude Desktop account. Follow the steps in Narmi’s Developer Docs to add the correct configuration block and replace the file path, API token, secret, and URL with your specific values.
Finally, restart Claude Desktop completely (quit and reopen) and start a new conversation. For example, try asking Claude, "Can you show me my account balance?" If everything is working correctly, Claude should be able to securely access your banking data and return a response.
Please see Narmi’s Developer Docs for troubleshooting information and tips.
Keep your API token and secret secure. Never share your claude_desktop_config.json file or commit it to version control systems.
The Narmi MCP banking integration is now available in beta for select Narmi partners, with general availability planned for Q3 2025. Financial institutions interested in offering their members AI-powered banking can contact Narmi to join the early access program.
