AI is changing how investment professionals access information, work with complex data and support investment decision-making. But putting it to work in investment management requires more than a model. It also needs data, calculations and context that hold up to scrutiny and controls over who can see it. 

Today, Addepar is announcing a new collaboration with Anthropic as part of Claude for Financial Advisors. Through Addepar MCP, a new connector for Claude, clients can ask Claude complex portfolio questions in plain language and get answers grounded in Addepar’s governed portfolio data. 

We spoke with Bob Pisani, Addepar’s Chief Technology Officer, and Kunal Gosar, Head of AI, about what it takes to bring reliable portfolio intelligence into an AI environment, what this new connection makes possible and how Addepar MCP fits into our broader AI strategy.

Bob, why does connecting with an AI experience like Claude matter for investment professionals?

Bob Pisani: AI is only as useful as the context behind it. In investment management, even seemingly simple questions often depend on a significant amount of underlying portfolio information. 

That’s where the Addepar MCP comes in. It gives Claude a secure way to call permission-aware Addepar tools and bring the resulting portfolio intelligence into the conversation. 

So a user could ask what drove a household’s performance, where it has exposure to a particular geography or whether it has enough liquidity to meet upcoming capital calls. Addepar does the hard work underneath: resolving the right portfolio context, applying the appropriate calculations, layering domain intelligence and returning the relevant results. Claude can then help the user explore and communicate those results naturally. 

What’s powerful about this is that clients don’t have to rebuild the portfolio intelligence layer themselves. They can bring Addepar’s data model, calculations and permissions into the AI environment they already use. 

Why build with Claude, and what does MCP make possible?

Bob: One of the things we like about MCP is that it’s an open standard. We believe firms should have flexibility in how they build their technology ecosystems and in the models, interfaces and experiences they choose to use. That’s consistent with how we’ve always built Addepar: as an open platform designed to work across clients’ broader technology ecosystems. 

Anthropic has been at the forefront of that approach with MCP, and Claude is good at the kind of multistep work our clients do every day. For firms already working in Claude, MCP creates a natural way to bring Addepar’s portfolio and financial intelligence directly into that experience. 

The important distinction is that Addepar is not simply supplying data for Claude to interpret. Addepar provides the intelligence behind the portfolio analysis. Our data model fully represents the complex relationships among families, households, accounts, entities, ownership structures and investments. Addepar resolves the relevant portfolio, applies the appropriate calculations and firm configurations, and enforces the user’s permissions. Through our connector, Claude can call these Addepar tools and receive results and actionable insights as opposed to just the underlying data.

Users can then explore follow-up questions, combine those results with other authorized sources and apply them across broader workflows, such as preparing analysis or drafting client communications. This gives clients more flexibility in where they work while Addepar remains the governed portfolio-intelligence foundation. 

Kunal, portfolio questions can look simple on the surface. What makes them difficult for a general-purpose AI system to answer well?

Kunal Gosar: The challenge is that the data itself is only part of the answer. You also need to understand how the assets, entities and relationships fit together and apply the right calculations, configurations and permissions. 

Take a household with assets across individual accounts, trusts, partnerships and private investment vehicles. If someone asks, “How did this household perform?”, you first have to determine what actually belongs to that household and how those entities relate to one another. From there, the system needs to apply the firm’s methodology and make sure the person asking the question is entitled to see the result. 

Addepar brings those pieces together because it already knows who owns what, how to calculate it, how public and private investments fit together, and who’s allowed to see the result. Because those capabilities are built specifically for investment management, Addepar can interpret the question in the right domain context and return a result that reflects how investment professionals actually work. 

Through the connector, Claude can call purpose-built Addepar tools for everything from simple portfolio queries to multistep analysis. The work happens inside Addepar, and Claude gets back the data it needs to answer the question.

Once that connection exists, how do Addepar-authored skills help turn it into something useful for clients?

Kunal: MCP gives AI applications access to the right tools. Skills help them understand how to use those tools well for a particular job or outcome. 

That matters because an investment workflow usually isn’t a single query. If you want to understand performance, for example, you may need to identify the right household, pull performance across public and private investments, find the biggest contributors and detractors and then decide how to present the answer. 

We can write those steps into an Addepar skill. So instead of requiring every user to know which tools to call or how the portfolio is structured underneath, we can give Claude a repeatable way to work through the problem based on our rich understanding of the domain, just like one of our experts would.

We’re starting with four areas where we think skills can be particularly useful: portfolio performance, performance attribution, total-portfolio exposure and private fund cash-flow analysis. We’ll continue to expand that set in partnership with clients as we learn where these skills can deliver the most value. 

What are some of the questions clients will be able to explore?

Kunal: We’re starting with questions where having a true total-portfolio view makes a meaningful difference. 

A user might ask: 

  • “What drove performance for this household last quarter?” 

  • “How much exposure does this portfolio have to emerging markets?”

  • “What capital calls are due in the next 90 days?” 

  • “Does the household have enough liquidity to cover its upcoming commitments?”

  • “Summarize this household’s performance, allocation and recent activity.” 

The emerging markets exposure example is a good illustration of why the underlying context matters. You may have direct exposure through public securities, indirect exposure through funds and additional exposure within private investments. Looking only at security names or top-level holdings won’t necessarily give you the full picture. 

Addepar can resolve that exposure across the portfolio using the information available to the firm and the user’s existing permissions. Claude then gives the user a natural way to explore the result, including asking the obvious follow-up questions. 

Data access and permissions are major considerations for firms adopting AI. How are those handled through the Addepar MCP connection?

Kunal: We designed the connection around the same identity and permissioning model clients already use in Addepar. 

When a user authenticates through their Addepar account, their existing permissions and firm context carry through to the tools our MCP exposes. Addepar still determines what portfolio information that user is authorized to access and performs the relevant calculations within our environment. 

That distinction is important. Claude is not given unrestricted access to a firm’s Addepar data. It receives the derived results needed to respond to the user’s request, subject to that user’s permissions. 

The initial experience is also deliberately focused on analysis and information retrieval. It does not execute transactions or alter portfolio data. And, as with any AI-assisted workflow, firms should apply their own review policies and controls to the outputs. 

Bob, how does the Addepar MCP complement Addison, Addepar’s purpose-built agents and ADX?

Bob: We think about all of these capabilities as part of the same underlying intelligence layer in Addepar. 

Addepar has spent years building the data model, calculations, ownership structures and workflows required to understand complex portfolios. AI lets us build on that foundation and make Addepar intelligence easier for clients to access and apply.

Addepar's clients leverage our capabilities in a variety of different ways. This launch empowers firms to access Addepar's portfolio intelligence through Claude for Financial Advisors, using Addepar MCP. This new offering complements the Addepar-native experience, which is what we’re building with Addison and our purpose-built agents. For clients with more complex data environments and requirements, they'll use ADX to bring Addepar data together with their own data, models and analytics. All of these capabilities across Addison, agents and ADX will be available via the Claude connector and the Addepar MCP.

The important thing is that clients don’t need to choose between flexibility and a trusted foundation. 

The AI ecosystem is going to keep evolving. There will be new models, new interfaces and new ways of working that we can’t predict today. Our strategy is to make sure Addepar’s intelligence can travel with our clients into those experiences while the portfolio foundation underneath remains consistent, governed and trusted. 

Learn more about the Addepar connector in Claude

The Addepar MCP connector is available today to eligible clients as part of its beta program. Addepar will work with participating firms to learn from real investment and operational workflows, refine the experience and shape how the capability evolves. 

Contact your Addepar representative to learn more about the Addepar MCP connector, availability and potential use cases for your firm.