Fund administration from your chat window

A few days ago I was onboarding a venture fund onto my fund admin app, collecting my exports of journal entries, spreadsheets, financial statements, K1s, and capital account reports to set up the limited partner register, schedule of investments, financial statements, and reconcile all of the historical data to make sure the ending state matched the data spread across different sources. It takes time to copy/paste and upload the documents and then review the journal entries, even with the built-in AI assistant. But then I had a thought: the app already had an MCP server to pull any data out of it, can't I use the MCP server to push all the data in through a Claude or ChatGPT chat?

Open AI and Claude have been racing to add interactions powered by their models into the surfaces we use to get things done (spreadsheets, documents, our computers, development environments, et. al.), and these surfaces have been working to integrate LLMs to make it easier for us to use their platforms. I built an AI-powered analyst into my open-sourced fund admin app to help people analyze information, enter in data, and provide additional context and analysis about the data in the app (portfolio company KPIs, fund performance metrics and benchmarks, and more) as well as run data extraction for portfolio company updates and run a due diligence process for a potential investment. ¹

But that still required me to go into the app to use all of it. Like many of you, I have found my workflows moving increasingly into chat, taking advantage of plugins, connectors, MCP servers, and MCP apps from services and software I use to get LLMs to help me get more ambitious work done faster. I wanted to enable the same thing with the fund admin app, so I spent a few days working to extend the existing MCP server into something that allows the user to not only read data from the app but also write data into the app.

Portfolio is now a headless open source fund administration app. "Headless" is a term meaning that the system keeps its data and logic separate from any one interface, enabling the user more choices in how to interact with the data and logic in the software. For portfolio tracking and fund administration, what that means is that this app can keep portfolio company performance, the schedule of investments, limited partner capital accounts and documents, and financials in one instance that you can run on your own infrastructure (including entirely local if you do not want to run it on the web) ², and reach it from a web browswer, Claude or ChatGPT over MCP, or through scripts over the API, all through authenticated, auditable, admin-configured user roles (CFO, GP, audit, tax, etc.).

• Setup the custom connector and authenticate the app with your LLM, and then you can open up your ChatGPT or Claude app on your phone and ask for information, charts, or reports in the app without logging into the app.
• Use the forecast feature to ask what would happen to your fund metrics if a portfolio company exited at a valuation.
• Email it new information about a portfolio company and ask it to update the latest KPIs.
• Run a due diligence checklist against a data room and get a list of findings and items to dig into.
• Setup a new investment by uploading the share purchase agreement and asking the app to create a new company.
• Create a capital call, distribution notice, fund performance report, and limited partner capital account statement via the web or directly in chat.
• Turn on the LP portal and LP MCP server if you want to give that same sort your LPs to access their documents via chat.

Do all of that in the app or with the app's AI analyst (load your desired LLM API key, Openrouter, Claude, and ChatGPT are prebuilt), or just do it in your ChatGPT or Claude chats to power all the AI features. Setup the app by cloning or downloading the software onto your computer, and then opening up a chat to point it to the folder and ask it to set it up. ³

And of course, it's open source, so if you would prefer it to look or work differently for you, change it to fit your desired workflow.

I think headless is an important innovation because it brings data into the surfaces you want to use. Less friction to get the data and insights you need, more flexibility in how you can use it, and greater ability to link it to other platforms you use, inside the places you use everyday, without having to learn a new app.

Does this remove the need for a fund admin? Definitely not. But for smaller entities that may not have robust reporting requirements - SPVs being monitored on spreadsheets, GP entities or management companies with little transaction history - the app replaces the manual work related to investment performance reporting, financial statements, capital accounts, and limited partner letters. For larger entities, the primary usecase is for portfolio performance tracking, KPI tracking and reporting, portfolio company performance and analysis, as well as an aggregation layer for performance reporting for entities that may be managed across multiple fund admins and platforms. For venture CFOs like me, managing financial and portfolio data across a variety of platforms is a daily challenge. Thee flexibility in use and deployment is designed for venture financial operations leads and builder CFOs that are looking to upgrade the financial operations of their firms. Whether you choose to use the software as-is, modify it extensively, or are inspired to build your own, it's a great time to build your own tools.

Questions, contact me anytime.