Wealth MCP How It Works | Moring
Model Context Protocol gives wealth firms a standard way to connect AI agents to firm systems, with each MCP server scoped by the identity minted from an advisor sign-in rather than anything the model sends. Firm permissions separate read access from action authority, gateways resolve identity and authorize every call, and governed deployments contain sensitive research and client data while capturing records that EU AI Act deployer duties and adviser recordkeeping rules expect.
On this page
Short answer
Model Context Protocol gives wealth firms a standard way to connect AI agents to firm systems, with each MCP server scoped by the identity minted from an advisor sign-in rather than anything the model sends. Firm permissions separate read access from action authority, gateways resolve identity and authorize every call, and governed deployments contain sensitive research and client data while capturing records that EU AI Act deployer duties and adviser recordkeeping rules expect.
How it works
Model Context Protocol, or MCP, is an open standard created by Anthropic that defines how AI applications connect to external data sources and tools, and wealth adoption guides describe it as USB-C for AI: a universal connector that lets any AI agent access any data source through a standardized interface without custom integration work for every combination. For wealth firms the appeal is immediate because advisor knowledge lives in systems never designed to be read together: CRM relationships, portfolio holdings, planning illustrations, enterprise documents, and research. MCP turns each of those systems into tools an agent can discover and invoke. The openness is the benefit and the risk in equal measure, which is why governed deployment is the whole conversation. Sources: https://www.milemarker.co/resources/guides/mcp-wealth-management
Permissions are the first wealth-specific concern. MCP gives applications a standard way to connect to external systems, and its tools specification describes operations that models can discover and invoke, but it does not by itself define a wealth firm approval policy; that responsibility belongs in the application, the connected systems, and the other controls the firm adds. Firms separate what a connection can read from what it can change: a connection retrieving a client briefing does not need the same authority as one updating a client record. Before enabling either, the firm establishes whose identity the connection uses, which records it can reach, which operations it can perform, and when a person must approve the result. Sources: https://www.onevest.com/blog/mcp-permissions-wealth-management
Identity scoping is enforced on every call rather than once at sign-in. One wealth-platform connector documents the discipline directly: the connector is scoped to the advisor who signed in, and which clients it can read is decided by the identity minted from that sign-in, never by anything the model sends, so no phrasing of a question can widen it. Two account properties are checked on every call, the advisor role and the MCP entitlement tied to the firm contract, with tools further split into read access and write access, and a refusal arrives as a sentence when either check fails. Identity from sign-in, entitlement from contract, and authority per tool travel together, which is the pattern a wealth compliance committee can specify in policy. Sources: https://developer.wealth.com/mcp/security
A gateway architecture is how firms industrialize those checks. A Model Context Protocol gateway is a protocol-authority host that sits between MCP clients such as agents, IDEs, and orchestrators and the downstream systems they want to reach, owning the protocol, resolving identity, authorizing every call, dispatching to a backend, and recording what happened, with cross-cutting concerns loaded as plugins across a stable boundary. For wealth firms, that means policy, redaction, and logging live in one auditable place rather than being reimplemented in each agent or server. A broker pattern generalizes the same idea across a portfolio of MCP servers, one per firm system, so connecting a new agent never means re-architecting the firm controls. Sources: https://mcpg.dev/docs/concepts/architecture https://ibl.ai/solutions/financial-services/mcp-guide
Financial-services MCP guidance turns the architecture into deployment steps. Practitioner walkthroughs for financial services describe inventorying the systems, building MCP servers for each, adding an MCP broker, connecting agents, and then running example workflows such as portfolio risk assessment across systems, regulatory compliance checks, and client onboarding due diligence, with security checks woven throughout. That sequencing matters for a wealth firm because it front-loads the permission and security conversation before any agent touches client data, instead of retrofitting control after the first integration demonstrates convenience. The broker also gives the firm one place to retire a vendor MCP server when a contract ends. Sources: https://ibl.ai/solutions/financial-services/mcp-guide
Records and information barriers are the buy-side extension of the same design. Investment-management MCP guidance describes one governed gateway attributing every tool call to an identity, containing material non-public information across pods, keeping positions and LP data out of external models, and capturing the records the adviser recordkeeping regime and the EU AI Act expect when agents connect to research stacks, deal rooms, order and execution systems, portfolio and risk systems, and the CRM holding investor records. Those systems carry exactly the information whose movement a wealth compliance officer must be able to explain. Sources: https://docs.mcpmanager.ai/industries/investment-management
The takeaway for wealth firms is that MCP is a permission surface, not just a convenience. The protocol supplies connection and discovery; the firm supplies identity binding, read and write separation, human approval points, and record capture through its gateway and policy layer. Where those are in place, an agent querying household data across systems is as supervisable as the advisor querying them directly, and the firm gains throughput without inventing a parallel audit process. Where they are absent, MCP becomes the fastest route to connecting agents to data no one cleared. Sources: https://www.onevest.com/blog/mcp-permissions-wealth-management https://docs.mcpmanager.ai/industries/investment-management - federalreserve.gov: https://www.federalreserve.gov/supervisionreg/srletters/SR2602.htm - federalreserve.gov: https://www.federalreserve.gov/supervisionreg/srletters/srletters.htm - artificialintelligenceact.eu: https://artificialintelligenceact.eu/high-level-summary/
Frequently asked questions
What is MCP in the context of wealth management?
Model Context Protocol is an open standard created by Anthropic defining how AI applications connect to external data sources and tools, and wealth guides describe it as USB-C for AI: a universal connector letting any agent reach any data source through a standardized interface without custom integrations for every pair. Wealth firms use it to give agents governed access to CRM relationships, portfolio holdings, planning illustrations, and research. The openness creates the value and the risk, so permission discipline around each connection is the central design conversation.
How should wealth firms scope MCP permissions?
Firms separate what a connection can read from what it can change, because a connection retrieving a client briefing does not need the authority to update a client record. Before enabling either, the firm establishes whose identity the connection uses, which records it can reach, which operations it can perform, and when a person must approve the result. MCP itself does not define approval policy; that belongs in the application, the connected systems, and surrounding controls, which is why wealth deployments layer policy on top of the protocol rather than assuming it.
How is advisor identity enforced on every MCP call?
The connector is scoped to the advisor who signed in, and which clients it can read is decided by the identity minted from that sign-in, never by anything the model sends, so no phrasing of a question can widen access. On every call the connector checks that the account holds the advisor role and the firm contract includes the MCP entitlement, with tools split further into read and write access; a refusal arrives as a plain sentence when either check fails. Identity, entitlement, and per-tool authority travel together on every request.
Why do gateways and brokers matter for MCP in regulated wealth deployments?
A Model Context Protocol gateway is a protocol-authority host that owns the protocol, resolves identity, authorizes every call, dispatches to backends, and records what happened, keeping cross-cutting policy, redaction, and logging in one auditable place rather than in each agent. Broker patterns add dedicated MCP servers one per system and connect agents through a governed intermediary. Buy-side guidance describes one governed gateway attributing every tool call to an identity, containing sensitive research and client data, and capturing records that EU AI Act deployer duties expect.