Using Claude to track and analyze your net worth with MCP

You can now give an AI assistant like Claude access to your financial data through something called an MCP connector. Ask "how concentrated is my portfolio" or "when do I reach my retirement number" in the same chat you already use, and get an answer computed from your own accounts. This is new enough that most people have not sorted out what these connectors actually do, or how they differ. Here is the map.

MCP (Model Context Protocol) is a standard way for an AI assistant to call an outside tool. A finance MCP connector is a small service that exposes your financial data, or computations over it, as tools the assistant can call during a conversation. You add the connector once, authenticate, and from then on the assistant can pull the data or run the analysis on demand.

The key thing: the assistant does not store your finances and the connector does not do the talking. The connector supplies structured, rule-computed facts, and the assistant turns them into an answer. That split matters when you compare options.

Almost every finance MCP falls into one of two camps, and they are solving different problems.

Read-only sync connectors. These connect to your banks and brokerages (often through an aggregator covering thousands of institutions) and expose live balances and transactions. Tools like Truthifi or a Monarch connector sit here. Strength: automatic, broad institution coverage. Limit: they hand the assistant raw balances, and the analysis is only as good as what the assistant improvises on the spot. They also inherit the sync blind spot, anything without an institutional login is invisible.

Analysis connectors. These focus less on pulling every balance automatically and more on computing decision-grade answers: net worth attribution, concentration and correlation, retirement projections, stress tests. The data can come in by hand or by import. Strength: the hard numbers are computed deterministically, not improvised. Limit: you do more of the data entry.

Neither is strictly better. If your whole financial life is US bank and brokerage accounts and you mainly want a tidy live balance, a sync connector is the easier fit. If you hold things that do not sync, or you want the assistant to answer forward-looking questions with real math behind them, an analysis connector does more.

A fair question: if the assistant is smart, why not paste your holdings into the chat and let it do the math? For a one-off question that works. It breaks down for anything you want to be right and repeatable.

An assistant asked to compute a portfolio's 90-day correlation or a Monte Carlo projection in-context will approximate, and will approximate differently next time. A connector runs the same computation the same way every time, on stored data, and hands back a number the assistant can trust and explain. The assistant is still doing what it is good at, judgment and language. The connector is doing what it is good at, deterministic computation and memory.

Opula is an analysis connector. You add it to Claude or ChatGPT, tell it what you own (including the assets no aggregator can sync), and it stores everything in a per-user database, converted to one currency. Then it answers with rule-computed reads: a market brief on your holdings, a wealth brief on net worth and cash flow, retirement projections, and what-if scenarios.

It is deliberately diagnostic. It reports what is true about your money and models what would happen under assumptions you supply, and it leaves the "so what should I do" to you and your own advisor. It is also built for the assets and currencies most US-first tools skip, which is the main reason to reach for it over a sync-only connector.

If you are choosing a finance connector, sort the shortlist by the two questions above: does it cover the assets you actually hold, and does it compute the answers you actually want, or just hand your balances to the model and hope.