Agentic AI
What is MCP?
Model Context Protocol
The Model Context Protocol is an open standard for connecting AI models to external tools and data over JSON-RPC. An MCP server exposes capabilities — queries, actions, resources — that a model can invoke through a uniform interface rather than a bespoke integration per application.
MCP matters for financial software because it makes the boundary explicit. A model does not receive an account; it receives a set of named capabilities, each with a schema, each individually authorised. That is a far easier surface to reason about than an agent holding credentials directly.
Where a capability moves money, the protocol boundary is the natural place to require human approval. The model can propose a transfer; the approval step belongs outside the model's control, which is what a passkey prompt on the owner's device achieves.
Related
Definitions are educational and general. Nothing here is financial, legal or tax advice, and regulatory descriptions summarise frameworks that change — check the current rules for your jurisdiction before relying on them.