Connect ChatGPT to your portfolio

ChatGPT supports remote MCP servers through custom connectors. The mechanics are close to the Claude lesson, with a few differences worth knowing before you start.

As before, the click-by-click steps live in the Connect Your AI documentation, because OpenAI moves these menus periodically.

What is different from Claude

Plan requirements. Custom connectors are not available on the free tier. They require a paid plan, and the exact tiers change — check the documentation page for current support.

Developer mode. Adding a custom connector generally sits behind a developer or advanced setting rather than the main connector list. It is a toggle, not a technical step.

Connector naming. You give the connection a label. Use something obvious like “Agni Folio” so you can identify it later.

Everything after that is the same, because the underlying protocol is identical. The server does not treat ChatGPT differently from Claude.

The connection

You add a custom connector with the server URL:

https://agnifolio.com/mcp

Then the same four steps: ChatGPT discovers the server, you are sent to an Agni Folio consent screen, you choose read or read-and-write, and you are returned connected.

The same guidance applies — start with read-only. It covers the great majority of what people want, and write access can be added later.

Verify before you trust it

Same first question as the previous lesson:

List my holdings and tell me my total net worth.

Compare to the app. Matching numbers mean the connection and the data are sound.

What ChatGPT is particularly good for

Because ChatGPT is where many people already do general research and drafting, the strongest use is combining portfolio data with everything else.

Using my actual holdings, write a one-page summary of my financial position I could take to an appointment with an accountant.

My portfolio is heavily weighted to one sector. Explain in plain language why that is risky, using my real numbers as the example.

Based on my current holdings and a target of financial independence in 15 years, what are the two changes with the largest effect?

The last is the sort of question that needs both your data and general reasoning — exactly the combination a connected assistant enables.

Safety, restated

The safety lesson drew a line between pasting data into a chat and connecting a permissioned tool. This is the second kind, and it is the safer one:

You approve the scopes
Read access cannot change anything
Write proposals still require confirmation inside Agni Folio
Access is revocable from Settings → Connected AI
Every call is logged

Which does not mean stop thinking. Review connected clients occasionally and remove any you no longer use.

Course complete

You started from a blank sheet and a subtraction. You now have:

A net worth figure you understand, with assets and liabilities properly sorted
A monthly habit for tracking it
A single view across scattered accounts
A realistic sense of what AI helps with and what it gets wrong
Clear rules on what never to share
A portfolio in one base currency, entered sensibly
The ability to ask questions about your real data
A target for financial independence
Your own AI assistant connected to your portfolio

That is more financial infrastructure than most people ever assemble.

What to do next

This month: run one net worth check and ask your assistant the three questions from Module 4.

This quarter: fill the gaps you skipped — the old pension, the policy you were unsure about, the loan balance you estimated.

This year: revisit your independence target with real spending data, and check whether you have passed the coast milestone.

Whenever: make sure someone you trust could find all of this if you could not explain it. It is the step everyone postpones and the one with the largest consequences.

Start putting it into practice at agnifolio.com — the core tracker, Astra and the MCP server are all free. Questions about the course or the product? Get in touch.