Finom connects business banking to Claude and ChatGPT via MCP
Finom, the European neobank serving more than 300,000 businesses, has launched a read-only Model Context Protocol (MCP) connection that lets customers query their business accounts in plain language from inside Claude or ChatGPT. The move positions Finom at an early but meaningful junction between AI assistant platforms and financial infrastructure, an intersection that is quietly redrawing the competitive map for business banking across Europe.
The integration is deliberately limited at launch. Customers authenticate with their existing Finom credentials, those credentials are never passed to the AI provider, and access is scoped to their existing Finom role. The connection is read-only: it cannot move money, issue cards, or alter account settings. A second phase, in which the assistant would prepare actions for the customer to confirm before execution, is planned but not yet live.
Why MCP, and why now
Finom's choice of the Model Context Protocol rather than a proprietary integration reflects an industry-wide shift in how software ecosystems are assembling themselves. MCP, originally developed by Anthropic, is fast becoming a de facto standard for connecting external data sources to AI assistants. By adopting it, Finom avoids building and maintaining a bespoke interface for every AI tool its customers already use, and positions itself to benefit as the protocol gains broader adoption.
Ivo Dimitrov, Chief AI Officer and co-founder of Finom, frames the decision in straightforwardly commercial terms: "Our competitors are moving there, but more importantly, so are our customers. People already run their work through ChatGPT, Claude and other AI assistants, and they build their own workflows there."
The compliance rationale for launching read-only first is equally pragmatic. Write access over financial accounts carries significant regulatory obligations under European frameworks, and Dimitrov is candid that Finom has no intention of rushing that phase. The company holds a European payment institution licence, meaning it operates under the same supervisory perimeter as licensed banks, not outside it.
The convergence angle: AI assistants as the new banking interface
For cross-sector investors and strategists, the more significant question is not whether Finom specifically succeeds, but what happens to business banking infrastructure when the primary interface shifts from a bank's own application to a general-purpose AI assistant. The neobank layer, already a disintermediation of legacy banks, faces its own disintermediation risk as AI assistants become the default workspace for small and medium-sized business operators.
This dynamic is not confined to fintech. The same MCP-driven logic is beginning to surface in enterprise resource planning, payroll, and supply-chain tooling, wherever structured business data can be exposed to a natural-language interface. The capital implication is that incumbents in any of these verticals who delay adoption risk ceding the customer relationship to whichever AI assistant platform the customer already works inside.
For sovereign and institutional investors building positions in European fintech infrastructure, the read-only launch is a signal worth tracking. The harder commercial test arrives when write access goes live and Finom must demonstrate that it can maintain European data-protection and payment-service-directive compliance while operating inside AI environments controlled by US hyperscalers. Regulatory alignment between EU AI Act obligations and MCP-based financial integrations remains an open question, one that will likely produce guidance, and opportunity, in the next 12 to 18 months.
The SME banking sector in Europe has seen consolidation pressure since 2023, with several challenger banks retreating from business accounts. Finom's move to embed itself inside AI workflows may be as much a retention strategy as a product one: if querying your bank balance becomes as natural as asking a question, switching costs quietly rise.