Finom has introduced an AI connection that lets European business customers query account data through Claude and ChatGPT, bringing transactions, invoices and accounting information into conversational interfaces without enabling payments or account changes.
Amsterdam-based financial platform Finom has launched Finom MCP, a new integration that allows business customers to access information from their accounts through AI assistants, including Claude and ChatGPT. Built on the open Model Context Protocol (MCP) standard, the connection enables users to ask questions about business finances in plain language rather than navigating multiple sections of their banking application.
The feature is included across Finom’s plans and is being rolled out across the European Union. Customers can query transactions, wallets, invoices, cards and accounting data, as well as request summaries of recent business activity.
The initial release is deliberately limited to read-only access. Users can retrieve and analyse information, but the AI assistants cannot move money, issue cards or change account settings through the integration. The approach reflects a cautious entry into AI-connected financial services, where data accessibility is advancing faster than the safeguards and regulatory arrangements needed for automated financial actions.
How Finom MCP connects business accounts to AI assistants
Finom MCP uses the Model Context Protocol, an open standard designed to connect AI applications with external tools and data sources.
Rather than copying financial information into a chatbot manually, customers can connect their Finom accounts to a supported AI assistant and ask questions about their business activity.
At launch, the integration covers several areas:
- Transactions: Retrieve information about recent account activity.
- Wallets: Query available wallet information.
- Invoices: Access invoice-related records.
- Cards: Retrieve information about business cards.
- Accounting data: Query relevant financial and bookkeeping information.
- Business summaries: Request a morning briefing covering recent financial activity.
For example, a business owner could ask an assistant to summarise recent transactions, identify invoice information or provide an overview of activity across the business account.
This could make routine financial administration more accessible, particularly for entrepreneurs and small businesses that do not have dedicated finance teams.
The integration does not replace Finom’s banking interface. The company says its application remains the authoritative source for official statements and reporting, while AI-generated responses should be treated as informational outputs that may contain errors.
Why the first release is read-only
Finom has chosen to restrict the initial integration to retrieving information rather than executing financial instructions.
Customers cannot use the connection to transfer funds, issue cards or change account settings. This separation is important because financial data access and the authority to initiate transactions carry different levels of risk.
A read-only assistant can help users understand account activity without being given the ability to alter the underlying financial position. By contrast, an assistant that can initiate payments or modify account details would require additional controls around authorisation, authentication and customer confirmation.
Kristjan Kaar, Finom’s chief product officer, acknowledged that the company was not the first financial platform to introduce a public MCP integration. He said the initial release focuses on read-only access, with additional capabilities intended to follow.
The company has not published a timeline for introducing write functions.
The phased approach allows Finom to test how customers use AI-assisted account queries before extending the integration to actions with direct financial consequences.
Customer consent and account security
Finom says customers retain control over the connection and must initiate the process from within their chosen AI assistant.
The setup redirects users to a Finom-hosted consent screen, where they authenticate using their existing Finom credentials. The AI provider does not receive those login credentials.
The integration also uses role-scoped access. This means the information available through the assistant is limited by the permissions already assigned to the customer within their Finom account.
Users can revoke the connection at any time.
These controls are intended to separate authentication, data access and the AI assistant’s ability to interpret information. The assistant can retrieve data that the customer has authorised it to access, but it does not automatically gain broader account privileges.
Ivo Dimitrov, Finom’s co-founder and chief AI officer, said the main engineering challenge was defining the boundaries around the connection rather than focusing solely on the sophistication of the AI model.
That distinction is important for financial applications. Even when an AI assistant can interpret information accurately, its access should remain limited to the permissions granted by the account holder.
However, read-only access does not eliminate every security or privacy risk. Financial information retrieved through an integration may still be processed by the AI provider under the customer’s separate agreement with that service. Customers therefore need to consider the provider’s data-handling policies, retention practices and applicable privacy terms before connecting financial accounts.
MCP is emerging as a connection standard for business software
Finom’s launch reflects the growing adoption of MCP as a way for AI assistants to interact with external applications.
The protocol provides a standardised method for connecting AI systems with tools and data sources. Instead of requiring every software provider to build a separate integration for each AI assistant, MCP can provide a common framework for exposing selected capabilities.
For business software providers, this could make it easier to connect accounting platforms, customer relationship management systems, payment services and other applications to conversational AI tools.
Finom’s own media FAQ identifies Qonto and Pennylane as companies that introduced public MCP integrations before it. Their early adoption illustrates how financial and business software providers are beginning to compete on their ability to make company data accessible through AI interfaces.
The wider opportunity is to reduce the time users spend switching between applications to retrieve information.
An entrepreneur could eventually use one AI assistant to review financial activity, check invoices, examine customer records and prepare operational summaries across multiple connected services. The usefulness of such a system would depend on the quality of integrations and the permissions governing each data source.
For Finom, adopting an open standard rather than relying exclusively on proprietary connections may also make it easier to extend the service to additional applications and product areas.
European regulation and the role of open banking
Finom operates within the European financial regulatory framework and holds an Electronic Money Institution authorisation issued by De Nederlandsche Bank.
Its integration sits within a broader environment shaped by the EU’s Payment Services Directive framework, including requirements governing access to payment account information through regulated channels.
PSD2 established rules for authorised third-party providers to access certain payment account data with customer consent. This helped create the foundation for open banking services across Europe.
Finom’s MCP connection adds an AI interface to account information made available through its underlying systems. The distinction is that MCP is a technical protocol for connecting applications, not a financial-services licence or a replacement for the regulatory permissions required to access and process account data.
The security and compliance of the overall service therefore depend on the underlying account-access arrangements, consent mechanisms, data processing and responsibilities of the parties involved.
The next stage of AI-connected finance will raise additional questions if assistants are allowed to initiate payments or make changes to financial accounts. Those capabilities would require stronger transaction controls, appropriate authentication and clear customer authorisation.
The separation between read-only access and payment execution gives Finom a way to introduce AI functionality without immediately extending the assistant’s authority over customers’ funds.
Finom’s expansion and European SME customer base
The launch comes as Finom continues to develop its financial platform for small and medium-sized businesses across Europe.
The company raised €115 million in a Series C funding round in June 2025, bringing its total funding to more than €300 million, according to the figures in the announcement.
Finom reports serving more than 300,000 businesses across Europe. Its services are aimed at entrepreneurs and SMEs seeking digital financial management tools.
The MCP integration could add another layer to that proposition by making account information accessible through conversational interfaces. Instead of requiring customers to navigate dashboards and reports, AI assistants can help retrieve and organise information through natural-language questions.
For small businesses, the potential benefit is time saved on routine administration. Finance teams could use AI tools to prepare summaries, review transactions and locate invoice information, although the accuracy of those outputs would still need to be checked against official records.
The commercial impact will depend on how frequently customers use the integration and whether it becomes part of their regular financial workflows.
What comes next for AI-powered business banking?
Finom’s current release provides a relatively controlled introduction to AI-assisted financial management. The assistant can retrieve information, but customers must continue using the Finom platform for actions involving money or account settings.
The company’s longer-term direction points towards a more interactive model in which an AI assistant could prepare an action for the customer to review and confirm.
For example, an assistant might eventually identify an outstanding invoice, prepare a payment instruction or organise a financial task for approval. Such a system would need to distinguish between recommendations, prepared actions and transactions that have actually been authorised and completed.
The transition from information retrieval to financial execution is likely to be one of the most important developments in AI-enabled banking. It also introduces greater operational and regulatory complexity.
Finom has not announced a timetable for the next phase or specified which write capabilities will be introduced first.
For now, its MCP launch provides European business customers with a new way to query financial information through Claude and ChatGPT. The broader significance is that business banking data is beginning to move into conversational AI environments, potentially changing how smaller companies manage everyday financial administration.

