In-house Build

InvestKit: a live product built with AI

InvestKit is a live web app at invest-kit.com (opens in a new tab) that keeps every broker and asset of a Ukrainian investor in one journal and prepares the annual tax declaration from it. A Nexera AI co-founder built it fully from scratch in under 2 weeks and runs it with the same AI method we sell: written requirements, specs, tests and a senior review of every change.

The challenge

Why InvestKit exists

Ukrainian investors declare investment income themselves every year. With accounts at several brokers, that means converting each trade, coupon and dividend at the National Bank rate for its date, matching lots FIFO and rebuilding a year of statements in a spreadsheet.

InvestKit turns that year of spreadsheet work into a journal that is always up to date, with the declaration generated from it.

What we built

What the product does

  • One journal for US stocks, government and private bonds, real-estate funds, deposits and cash.
  • Statement import for Interactive Brokers, Inzhur and generic CSV. Re-importing a file never duplicates trades.
  • A tax module that calculates every lot at historical NBU rates and fills the F1 appendix and tax-office XML in the browser.
  • Analytics: value against benchmarks, XIRR / TWR, allocation and a 12-month income forecast.
  • Security built in: Google sign-in, Postgres row-level security, EU-hosted data and an append-only audit log.
Inside the product

InvestKit screens

Screenshots from the public demo, which runs on a made-up portfolio. You can open the same screens yourself.

InvestKit overview: a list of items that need attention, net worth of $12,480 with a growth chart, and a 9.40% XIRR return split into income, market and fees.

Overview. Net worth, return and what needs attention, such as an unfiled declaration or a coupon due soon.1 / 5

How AI was used

Built with the method we sell

InvestKit went through the same loop as every client project. AI did the heavy lifting at each step, and a senior developer signed off each gate.

  1. Requirements

    Every feature started as written requirements in plain language: what the journal stores, how a broker statement maps to trades, which tax rules apply.

  2. Specs

    Each requirement became a technical spec that the AI implemented against, such as lot matching by FIFO or converting each payment at the National Bank rate for its date.

  3. Tests

    Automated tests were written alongside the code, covering rules such as re-importing a statement never duplicating trades.

  4. Review

    A senior developer read and approved every AI-written change before it merged, including security-critical parts such as row-level security in Postgres.

By the numbers

What the method left behind

Counted from the InvestKit codebase in October 2026.

  • 2,681 automated tests

    All pass before a change merges. They cover tax rules, lot matching, statement import and access control.

  • 28 decision records

    Each important technical choice is written down with its reasons, so the next change builds on it.

  • 25 database migrations

    The data model grew in small, reviewed steps while the product stayed live.

  • 2 languages

    Every screen works in Ukrainian and English.

Stack

Technology

Mainstream, widely supported tools, the same kind we pick for client products.

  • Next.js
  • Supabase · Postgres
  • Google sign-in
  • Vercel
  • LiqPay
Outcome

Where InvestKit is today

Live at invest-kit.com (opens in a new tab) in Ukrainian and English, with a public demo, a free plan and a Pro tier. You can open the public demo and judge the product yourself.

Most of our work for other companies is under NDA. On a Scope Call we're happy to walk you through relevant experience, anonymised.

FAQ

Questions, answered

Is InvestKit a client project?

No. InvestKit is an In-house Build: a Nexera AI co-founder built it and runs it as a real product with a free plan and a Pro tier. It shows the method we use on client work.

Can I try InvestKit?

Yes. It's live at invest-kit.com (opens in a new tab) in Ukrainian and English, with a public demo and a free plan.

What was AI used for when building InvestKit?

AI wrote most of the code against written specs, wrote tests alongside it and summarised every change for review. A senior developer approved each change before it merged.

Can Nexera AI build a similar product for me?

Yes. A fintech or data-heavy web app like this usually starts as a Clickable Prototype for $100–300, then an MVP Launch from $4,500 in 2–4 weeks.

Is user data in InvestKit secure?

InvestKit uses Google sign-in, Postgres row-level security so each user only sees their own data, EU-hosted data and an append-only audit log.

Want a product like this?

Start with a free 30-minute Scope Call with a senior developer. You leave with Product Requirements and a Ballpark.

Last updated: