MCP: What It Is and Why It Matters for Family Offices

Last updated: August 25, 2026.

Quick Answer

MCP is a way for a family office to use its wealth data in AI tools while the data itself stays safely where it already lives. MCP (Model Context Protocol) is an open standard, a universal connector between AI assistants and the systems that hold your data. There is nothing to install and no new tool to learn. A family office platform with an MCP layer, such as Aleta, lets the office ask questions through its preferred AI assistant in plain language while keeping full control of its data.

Key Takeaways

• MCP (Model Context Protocol) is an open standard introduced by Anthropic in November 2024. OpenAI adopted it in March 2025, Google DeepMind followed in April 2025, and the standard now sits under the Linux Foundation’s Agentic AI Foundation and is not controlled by a single vendor.

• An API is an interface that connects software systems. MCP is an open protocol on top of an API that gives AI agents structured, permissioned access to data.

• 22% of family offices use AI for operational tasks or investment analysis, up from 13% in 2024 (Citi, AI in the Family Office, 2026).

• 16% of family offices use AI for investment performance reporting, more than double the share a year earlier (Citi, AI in the Family Office, 2026).

• Access through an MCP layer is permissioned. An AI agent works with your permissions: it sees the data you can see, and nothing more, under the same security standards as the platform itself.

What Is MCP in Simple Terms?

If you have ever opened Claude and added a connection to another tool you use, you already know what MCP does. The moment the assistant can suddenly see your calendar, your data in another tool, your files and work with them, that is MCP at work: the connection runs on it.

Connecting to a wealth platform works the same way. Your wealth data stays stored where it already lives, in your platform, governed and permissioned. Your AI tools come to the data and ask, and only the answer travels back. You keep control of the data at the platform level; what an assistant does with an answer depends on your agreement with that provider, which the FAQ below covers. In family office software, MCP is the open standard that makes this way of working possible.

The rest of this article is the fuller explanation: where the standard comes from, how it differs from an API, and what to check before connecting anything.

If you would rather start with results, our companion article Real Prompts You Can Steal: How Family Offices Use AI Agents on Their Wealth Data shows 11 real prompts and the full answers they returned in Claude.

What Is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard that defines how AI models connect to external tools, systems, and data sources.

Anthropic introduced MCP in November 2024. OpenAI adopted the standard in March 2025, and Google DeepMind followed in April 2025. The full specification and documentation are open and free for any provider to implement.

In December 2025, Anthropic donated MCP to the Agentic AI Foundation, a directed fund under the Linux Foundation co-founded by Anthropic, Block, and OpenAI. In plain terms, the donation means no single AI vendor controls the standard, its roadmap, or access to it. For a family office evaluating software, building on MCP is a commitment to an industry standard, and the office can change AI assistants without rebuilding its connections.

The most common analogy is USB-C. Before a universal connector existed, every device needed its own cable, and every combination of laptop and accessory was a separate engineering problem. MCP plays the same role between AI models and data sources. A system that offers an MCP connection can be queried by any AI assistant that speaks the protocol, and the integration work happens once, at the standard level. Each new assistant or data source that adopts the protocol plugs into everything already on it.

One more term matters before going further. An AI agent is an autonomous AI system that completes tasks with minimal human intervention, gathering data, running checks, and drafting output along the way. Ken Gamskjaer, Aleta’s CEO and co-founder, explores the concept in depth in The Rise of AI Agents in Wealth Management.

Family offices sit on unusually fragmented data: custodians, banks, fund administrators, general partners, property managers, and internal spreadsheets. AI adoption in the segment is still early. 22% of family offices use AI for operational tasks or investment analysis, up from 13% in 2024, according to Citi’s 2026 report AI in the Family Office. The same report finds that the offices moving first start with everyday work such as document summaries and report automation. MCP matters here because it turns those early experiments into governed workflows connected to real wealth data.

How Does MCP Differ From an API?

The two solve different problems, and a modern wealth platform needs both.

An API (application programming interface) is an interface that lets one software system request data from another: feeding a general ledger, synchronizing a CRM, or streaming positions into a BI tool. Working with an API is a developer task. Each integration is designed, coded, and maintained.

MCP is an open protocol that sits on top of an API and is built for AI agents. An API integration is coded in advance for one defined purpose. An MCP connection lets an agent discover what data is available and query it in a structured, permissioned way at the moment a question is asked.

The relationship is layered. Aleta, for example, exposes wealth data through an open API, with an MCP layer on top for AI agents. The API makes the data portable and integrable. The MCP layer lets AI agents query it securely and in a structured way. For a deeper look at how the two work together in practice, read the interview-style walkthrough Aleta API & MCP: Building a Future-Proof Tech Stack.

For human analysts who want to explore data without writing code, Aleta offers a third access route: the Data Cube. It stands on its own, separate from the API, and gives power users a live, no-code connection to Excel, Power BI, or Tableau. The rule of thumb: integrations and data pipelines use the API, ad hoc analysis and custom BI reporting use the Data Cube, and AI agents use the MCP layer. For a closer look at what the Data Cube does, read the companion interview with Aleta’s CEO & Co-founder, Ken Gamskjaer, The Aleta Data Cube: A Custom Reporting Superpower.

The table below shows how the three routes divide the work.

What Does Permissioned AI Access Mean for a Family Office?

Citi's 2026 report puts data privacy first among its key takeaways and notes that AI solutions unable to guarantee data security are unlikely to be adopted by family offices. The concern is well founded. When staff paste holdings into a consumer chatbot, the office creates an uncontrolled copy of sensitive data outside its security perimeter, with no audit trail and no way to revoke it.

Permissioned access exists to close exactly that gap. The assistant never holds a copy of the data. It holds access, and access can be audit-logged and revoked. You decide what an AI agent can see and what it can touch, at two levels.

The first control level is the platform. When you connect an AI assistant to Aleta through the MCP layer, the assistant works with your permissions: it sees the data you can see, and nothing more. The agent sends a structured query, the platform checks those permissions, and only the approved slice of data comes back for that specific question. The data stays in one place, under the office's control, and the AI tools come to it.

The assistant itself is the second control level. In Claude, for example, you decide which tasks it may perform on its own and which require your approval first.

Permissions govern what an assistant can see. What happens to an answer after it is delivered is a separate question, and no platform can control that. Responsibility there sits with the office and its chosen AI provider, so verify that your data is not shared with anyone, not used for training, and not retained. Enterprise AI deployments typically offer these guarantees contractually, and the FAQ below covers what to confirm in writing.

Aleta’s MCP layer applies this model on top of its permission-based API and is built with the same encryption and security standards that protect the core platform, which is SOC 2 Type II certified.

What Can a Family Office Actually Do With MCP?

The strongest early use cases remove recurring work from small teams.

Citi’s 2026 report finds the most successful AI applications in family offices solve everyday problems: summarizing documents, transcribing meetings, and automating reports. The same survey shows AI use in investment performance reporting more than doubling in a year, to 16% of family offices. MCP is the mechanism that lets this kind of work run against live, permissioned wealth data. Three patterns come up again and again in my client conversations.

Can an AI Agent Prepare a Family Office Morning Briefing?

An agent the office runs every Monday morning queries positions, cash, and recent transactions through the MCP layer and delivers a short briefing: what moved, what settled, what needs attention. Work that used to open the week with an hour of portal logins becomes a two-minute read.

Can AI Answer a Principal’s What-If Question in Minutes?

Principals rarely ask questions in report format. A question such as “what happens to our liquidity if we fund the next capital call from the operating entity” spans custodians, entities, and currencies. An agent with permissioned MCP access pulls the relevant balances and commitments and answers in plain language, with every number traceable back to the platform.

Can AI Summarize Private Markets Activity Across Entities?

Capital calls, distributions, and NAV updates arrive continuously and land in different entities. An agent can summarize the month’s activity, flag unfunded commitments, and draft commentary for the investment committee. On Aleta, this builds on data that Aleta Intelligence, the platform’s suite of AI tools, has already read, mapped, and booked from the underlying documents.

Curious what this looks like with real prompts and real outputs? Read Real Prompts You Can Steal: How Family Offices Use AI Agents on Their Wealth Data in our knowledge hub.

Why Does MCP Depend on a Structured Data Foundation?

An AI agent inherits the quality of the data it queries.

You cannot build on top of noise. If positions are stale, entities are mismatched, or private markets documents sit unprocessed in inboxes, an agent will produce fluent answers with wrong numbers, faster than any human would. The value of MCP access rises and falls with the reconciliation discipline underneath it.

This is why MCP support signals more than a feature checkbox. An MCP layer earns trust when it sits on a structured, machine-readable data foundation. At Aleta, a dedicated Data Management Team reconciles bank and custodian feeds daily through data reconciliation workflows, so agents query verified data. Ken Gamskjaer, Aleta’s CEO and co-founder, makes the fuller argument in this piece: fix your data before you buy AI.

Citi finds that 57% of family offices cite lack of internal expertise as the biggest barrier to AI adoption. A protocol-level standard lowers that barrier. The office grants access and asks questions. Integration work stays at the protocol level, maintained by the vendors who adopted it.

Which Family Office Platforms Support MCP Today?

Announcements about AI features now come from every corner of the family office software market. MCP support is only now beginning to appear.

As of August 2026, based on public documentation, the market is splitting into two camps. Addepar, Masttro, and Eton Solutions document AI assistants that operate inside their own products (Addison, Masttro Intelligence, and EtonAI), with no published MCP connection. A first group of platforms has begun opening up: Aleta offers native MCP support as part of its open architecture, Asora has launched a read-only AI connector built on MCP, and d1g1t, a platform serving advisors and multi-family offices, released an MCP server in July 2026. Aleta remains among the first family office platforms with native MCP support, and the fact that others are following is a healthy sign for buyers: MCP support is on its way to becoming a standard expectation.

A closed platform can add AI inside its own walls, and the vendor selects the models, the use cases, and the pace. An open platform with an MCP layer hands those choices to the office. The office picks its own AI assistants, connects them to its own data under its own permissions, and sets its own agenda. The distinction runs deeper than AI, and it is worth understanding before any platform decision: open vs. closed architecture in wealth reporting covers it in full.

As MCP connections spread, the meaningful differences move down the stack. An agent inherits the data beneath it: how many asset classes are reconciled, whether private markets documents are read and booked, and how fresh the custodian feeds are. Two platforms can both speak the same protocol and still give an agent very different material to work with. That returns any evaluation to the data foundation covered above, and to the question of who sets the AI agenda for the office’s data.

What Should Family Offices Ask Vendors About MCP?

The vocabulary of open architecture is now universal in wealth technology, but readiness behind the vocabulary varies widely. Six questions surface the difference.

1. Is MCP support live today? A roadmap answer means AI agents cannot reach the data yet.

2. Does the MCP layer sit on an open, documented API? An MCP layer inherits the depth of the API underneath it.

3. How is access controlled? A connected assistant should work within the user's existing permissions, seeing what that user can see and nothing more. Ask whether every query is audit-logged, whether access can be revoked at any time, and whether anything beyond read access requires explicit approval.

4. Is client data used to train the vendor’s AI models? The answer should be an unqualified no, in writing.

5. Which AI assistants have been tested with the layer? Named assistants and reference workflows show the capability has shipped.

6. What data foundation sits underneath? Ask how documents are ingested and how feeds are reconciled before an agent can query them.

These questions tend to sort the market quickly. Aleta, named Best Data Provider at the Family Wealth Report Awards 2026, offers MCP support today as part of its open architecture.

The pattern behind all six questions is ownership. Aleta’s position here is long-held: a family office should own its data outright, route it into a best-of-breed stack of specialist tools, and set a technology agenda independent of any single vendor’s roadmap. An MCP layer is that belief applied to AI. The office that owns its data can ground AI agents in its actual holdings, documents, workflows, and permissions. Ownership of the data becomes ownership of the AI agenda.

I made the broader argument in the July edition of the Aleta Edge newsletter: the mid-year truth about family office technology and data ownership.