# Things I've built

Some of this was built for insurers. Some of it I built because I wanted the tool to exist. Each page says what the problem was and how it was solved.

## Nola

Booking and quoting software for local service providers: salons, shops, studios and solo chairs.

- Category: products
- Status: 2026
- Tags: SaaS, Booking, Payments
- Site: https://getnola.co

### The problem

Small service businesses run on a phone, a calendar and a pile of messages. Fixed-price bookings and custom jobs that need a quote live in different places, and neither is connected to payments or who actually shows up.

### How it's built

- A provider page that goes live with services, durations, prices, hours and staff — customers can only book slots you opened.
- Custom jobs arrive as quote requests with details attached and turn into bookings when the customer accepts.
- One day view with every resource and staff member side by side, so double bookings and gaps are obvious.
- Prepaid wallets: customers top up a balance and spend it on bookings.
- Door access and lights tied to the booked hour instead of a shared code.

### Links

- [getnola.co](https://getnola.co)

[read more](https://kivilaid.ee/projects/nola)

## Nora — Insly AI

An AI layer that removes manual work from insurance operations, so MGAs and insurers scale without adding headcount.

- Category: insurance
- Status: live
- Tags: AI agents, Underwriting, MGA
- Site: https://ai.insly.com

### The problem

Broker submissions arrive in inconsistent formats, underwriters lose hours chasing missing information, and every referral means opening the underwriting guide and searching for the right section by hand.

### How it's built

- Extracts data from broker submissions and direct digital submissions, whatever the format.
- Maps the extracted data onto the insurer's own templates.
- Chases brokers for missing information automatically.
- Exports finished submissions to the policy administration system.
- Answers questions about the underwriting guide and cites the exact section the answer came from.
- Connects via API or works standalone.

### Links

- [ai.insly.com](https://ai.insly.com)

[read more](https://kivilaid.ee/projects/nora)

## Eesti Data MCP

One MCP endpoint for Estonian public data — 77 tools, no API keys.

- Category: public-data
- Status: live
- Tags: MCP, Open data, Claude Code
- Site: https://eesti-data.myservice.ee/mcp

### The problem

Estonian public data is excellent but scattered across dozens of registers and APIs, each with its own format. Asking an AI assistant something simple — today's electricity price, who owns a company, how the Riigikogu voted — meant stitching sources together by hand.

### How it's built

- A single remote MCP server with 77 tools across energy, weather, companies and financials, taxes, parliament, maps and cadastre, transport, legislation, statistics, schools, nature and more.
- Public HTTP endpoint — no authentication, no API keys.
- Ships with a Claude Code skill, so the assistant knows which tool fits the question.
- Operated as a live service rather than a repo you have to host yourself.

### What it does

- 77 tools
- 15 data domains
- 0 API keys needed

### Install in Claude Code

```bash
claude mcp add --transport http eesti-data https://eesti-data.myservice.ee/mcp
```

### Try asking

- "What is the electricity spot price in Estonia right now?"
- "How did the Riigikogu vote last Wednesday?"

### Links

- [eesti-data.myservice.ee/mcp](https://eesti-data.myservice.ee/mcp)

[read more](https://kivilaid.ee/projects/mcp-servers)

## AI Visibility Tracker

Tracks whether ChatGPT, Claude, Gemini, Perplexity and other AI answer engines mention and recommend your business.

- Category: products
- Status: 2026
- Tags: GEO, Next.js, Convex, AI SDK
- Site: https://ai-visibility-tracker-sooty.vercel.app

### The problem

More and more buyers ask an AI assistant instead of searching. Most businesses have no idea whether those assistants know they exist, who gets recommended instead, or what to fix.

### How it's built

- Multi-tenant SaaS built on Next.js, Convex and the Vercel AI SDK.
- Detect: daily tracking of mentions and citations per engine and per prompt.
- Diagnose: a 13-category GEO audit of every page for AI-crawler readiness.
- Produce: one-click content briefs, llms.txt and JSON-LD fixes generated from what's missing.
- Verify and prove: re-audit diffs and correlation with real AI-referral traffic.

### What it does

- 8 AI engines tracked
- 13 audit categories

### Links

- [ai-visibility-tracker-sooty.vercel.app](https://ai-visibility-tracker-sooty.vercel.app)

[read more](https://kivilaid.ee/projects/ai-visibility-tracker)

## Hangete Radar

AI-powered discovery of Estonian public procurement: summaries, smart matching and market analysis.

- Category: public-data
- Status: 2026
- Tags: Public procurement, Claude, Vector search
- Site: https://hanked.vercel.app

### The problem

Estonian public tenders are published in the national register, but finding the relevant ones means browsing CPV codes by hand and reading 100-page PDFs to learn the requirements and deadlines.

### How it's built

- Tender data synced directly from the national procurement register.
- Faceted search by value, location and a browsable CPV code tree.
- Claude-generated summaries of requirements and deadlines, in Estonian.
- Vector-based alerts that match new tenders to a company's profile.
- Market analysis of buyer spending patterns and competitors' wins.

### What it does

- 40+ filter types
- 9k+ CPV codes

### Links

- [hanked.vercel.app](https://hanked.vercel.app)

[read more](https://kivilaid.ee/projects/hangete-radar)

## YouTube Insight AI

Paste a YouTube link and get a detailed summary and key insights, generated by Gemini.

- Category: products
- Status: 2026
- Tags: Gemini, Convex, Clerk
- Site: https://you-tube-insight-ai.vercel.app

### The problem

Long videos hide a few minutes of value. Watching them to find out is slow, and generic summarisers lose the structure and the specifics.

### How it's built

- Vite and React front end with a brutalist editorial design.
- Clerk for authentication and billing, Convex for generations, entitlements and RAG.
- A free tier on a hosted key with monthly budgets; paid plans unlock private generations and custom prompts.
- Bring-your-own Gemini key, stored encrypted, for unlimited use.
- A remote MCP endpoint, so AI assistants can use it as a tool.

### Links

- [you-tube-insight-ai.vercel.app](https://you-tube-insight-ai.vercel.app)

[read more](https://kivilaid.ee/projects/youtube-insight-ai)

## FitBox

Turns a 3D scan of a space into the largest buildable plywood subwoofer enclosure, with cut patterns, bevel angles and predicted response.

- Category: tools
- Status: 2026
- Tags: 3D geometry, Acoustics, In-browser solver
- Site: https://fitbox-blond.vercel.app

### The problem

Fitting a subwoofer box into an odd-shaped space — a car boot, a cabinet — usually means cardboard mock-ups, guesswork and wasted plywood.

### How it's built

- Upload an STL or OBJ scan; it's parsed and repaired in the browser and never leaves your machine.
- Calibrate scale, level the floor, close the cavity and mark keep-out zones.
- Pick a driver from the database; target volume, tuning and port length come from its Thiele-Small parameters.
- A deterministic solver finds the biggest rectangular, convex or non-convex box that respects every build constraint.
- Exports nested DXF and SVG, a cut list with bevel angles, 1:1 templates, STL, bill of materials and cost.

### What it does

- < 60 s typical solve
- ±1 mm verified fit
- 28 drivers in the database
- 6 export formats

### Links

- [fitbox-blond.vercel.app](https://fitbox-blond.vercel.app)

[read more](https://kivilaid.ee/projects/fitbox)

## Brand Scraper

Extracts logos, colours, social links and contact info from any company website.

- Category: tools
- Status: 2026
- Tags: Headless browser, API, Brand data
- Site: https://brand-scraper-andok.vercel.app

### The problem

Getting a company's logo, brand colours and contact details for a CRM, a pitch or an onboarding flow means clicking through their site and copying things by hand.

### How it's built

- A headless browser visits the site and collects meta tags and assets.
- Logos, favicons and backdrop images are extracted with their resolution.
- Brand colours come from CSS, theme tags and analysis of the logo.
- Social profiles, careers page and contact info are returned as structured data.
- Available as a self-hosted API with a playground and an AI query builder.

### Links

- [brand-scraper-andok.vercel.app](https://brand-scraper-andok.vercel.app)

[read more](https://kivilaid.ee/projects/brand-scraper)

## ChromaLens

Deconstructs any website's design system — colours, CSS variables and fonts — and has an AI critique the palette and layout.

- Category: tools
- Status: 2026
- Tags: Design audit, Web scraping, LLM critique
- Site: https://chromalens-ai-audit.vercel.app

### The problem

Auditing a site's visual design by hand means opening dev tools page by page to collect colours and fonts, and the critique that follows is mostly opinion.

### How it's built

- Discovers the sitemap and picks up to 10 representative pages.
- Fetches linked stylesheets and extracts real colour tokens, CSS variables and font families.
- Passes everything to a design audit model for palette analysis and layout critique.

### Links

- [chromalens-ai-audit.vercel.app](https://chromalens-ai-audit.vercel.app)

[read more](https://kivilaid.ee/projects/chromalens)


---

Source: https://kivilaid.ee/projects · Contact: https://www.linkedin.com/in/andokivilaid · Overview: https://kivilaid.ee/llms.txt