Most GPs don't just have a data problem. They have a distance problem: the information they need to answer a partner's question in a Monday meeting exists somewhere in the platform, but getting to it means clicking into fund after fund, company after company, and stitching the answer together by hand.
One of our clients decided to close that distance themselves. Using Claude and Atominvest's API, their team built a custom internal dashboard - an insights hub, if you will - that pulls live portfolio data straight out of Atominvest and reshapes it into the exact views their investment team actually thinks in.
What they built
The insights hub is a purpose-built layer on top of Atominvest, assembled around a few core modules:
- A fund and vehicle switcher. Click into any fund or co-investment vehicle and get an immediate, consolidated view of every portfolio company inside it, with figures displayed in whichever currency the viewer needs.
- Trend views. A rolling look at how a fund has performed over recent months, without having to reconstruct the timeline manually.
- Forecasting. Projected performance for the fund and for each underlying portfolio company individually - the kind of forward-looking equity story that used to live in a spreadsheet someone rebuilt every quarter.
- Health scores. A simple, visual signal for portfolio health at a glance, across companies and vehicles.
- Company-level deep dives. Highlights, lowlights, side-by-side comparison tables, and individual equity stories, with the ability to attach relevant slide decks directly to a company's view.
It's the kind of tool an internal analytics team might spend two quarters scoping, staffing, and building. This team built it in a fraction of that time, without adding headcount.
How they built it - and why the approach changed
The first version was built as an artifact connected through Claude's MCP integration, calling Atominvest's data live. It worked, but the team found it slower going and noticed the token cost of refreshing data on every interaction added up. So they rebuilt the backend to call Atominvest's API directly - a more stable, deterministic foundation with no risk of the model inventing numbers - and used Claude to design and iterate on the interface on top of it.
The result is a lightweight HTML application, deployed via GitHub and Railway, with portfolio data cached rather than fetched live on every view - which is also simply how the underlying reporting cycle works, since the data itself is only refreshed when it's uploaded. That caching choice is what makes the insights hub feel instant to the people using it day to day.
What's next
The team's already scoping their next module: a consolidated cap table view across every portfolio company - shareholders, ownership stake, and how that's shifted over time. It's a good example of the underlying pattern. Comparing three companies' cap tables side by side, or tracking a stake's movement across financing rounds, is possible today, but it means navigating into each company one at a time. A dedicated module would turn that into a single click.
Why this matters beyond one firm
This isn't really a story about one dashboard. It's a story about what happens when a platform's data is easily accessible: teams stop asking their vendor to build every view they can imagine, and start building the ones they actually need, on their own timeline. Atominvest's API - and its MCP integration for teams who want to prototype directly against Claude - exists precisely so that GPs aren't limited to the views we ship by default.
If your team has an idea for a module like this, the constraint is almost never the tooling. It's knowing which data points you need. That's a conversation worth having.
