How we built the first version of GeoUp in two months
Key takeaways
- GeoUp is our own SaaS product for AI visibility. Two senior developers built the first version in two months, from zero to production.
- Even the first version comprised around 250,000 lines of code and queried four AI providers with 30 industry-specific questions. On top of that came tools for improvement, billing via Stripe, admin dashboards and scheduled workflows.
- The speed comes from a clear division of labor: experienced developers plan the architecture and data model and review every result, while AI handles the routine work.
- Since the first version, GeoUp has grown with feedback from its users: a fifth AI system, features for agencies, reports, a new design and an MCP server.
As of October 2026
You are planning your own SaaS product, meaning software that your customers use online and pay for on a recurring basis (software as a service). Before you set a budget and a schedule, you want to know what is realistic within a few months. We show it using our own product GeoUp: what went into the first version, how we approached it and how the product has grown since.
It started with a question
Anyone who asks ChatGPT for the best Italian restaurant in Frankfurt gets a concrete recommendation. There was no Google Analytics for answers like that. Businesses could not see whether AI systems mentioned them, recommended a competitor or left them out entirely. David Sroka, who is responsible for product and engineering at njudev, describes the starting point like this:
"Everyone was optimizing for Google. We asked: what about ChatGPT? What about Claude? GeoUp is the answer to a question the market hadn't even formulated yet."
This question turned into a clear goal for the first version: measure how visible a business is in AI answers, and provide the tools to improve that visibility.
What went into the first version
The core was the measurement. GeoUp queried four AI providers, namely ChatGPT, Claude, Gemini and Perplexity, with 30 industry-specific questions across three scopes: city, region and nationwide. The platform analyzed every answer: Is the business mentioned, in which position, and which web addresses does the AI cite as sources? Monthly scans showed how visibility developed over time.
The tools for improvement built on this measurement:
- Authority Boost published company profiles on nine high-authority platforms.
- AI-generated landing pages provided structured data based on Schema.org, a shared vocabulary (schemas) for structured data on the web, founded by Google, Microsoft, Yahoo and Yandex.
- Market Battle compared a business's own visibility with that of its strongest competitor.
- Auto-Blog automatically created content for search engine optimization (SEO).
- Georgie, a built-in AI assistant with a voice, guided users through the entire process.
The platform was built with Next.js and TypeScript, a widely used framework for web applications and its associated programming language. The data is stored in a MySQL database, connected via the database interface Prisma. On top of that come the AI services of OpenAI, Anthropic, Google Gemini and Perplexity, the payment provider Stripe, the cache Redis, the animation library GSAP and the speech service Google Cloud TTS.
The first version comprised around 250,000 lines of code, four AI providers and nine publishing platforms. You can also find the key facts on our case study page.
How two months were possible
Two senior developers built the first version in two months. What makes this possible is the way of working we describe on our homepage under "How we work": architecture, logic and data model are settled before the AI begins. After that, humans and AI work together, and senior developers review, test and approve every result. Our article From idea to finished platform shows what this process looks like in a client project.
We have summarized what studies show about how much AI speeds up development in our article What does your own platform cost? According to these studies, AI has the strongest effect on new projects and clearly defined tasks. That is exactly how GeoUp came about: built from scratch, with a clear goal.
Our standard: a project that would traditionally take six months, we deliver in eight weeks. At two months, the first version of GeoUp falls squarely within that range.
Billing, administration and automation were part of it from the start
A SaaS product is a business from its first customer. That is why the first version of GeoUp included not only what users see, but also everything that keeps it running:
- Billing: It runs on Stripe Billing, the payment provider Stripe's service for recurring payments and subscriptions.
- Admin dashboards: The team keeps track of customers, workflows and results in dedicated overviews.
- Cron automation: Scheduled workflows take care of recurring tasks on their own, without anyone having to trigger them.
If you plan these building blocks from the start, you can onboard and bill customers from day one. The team can then focus on the product instead of manual work.
In practice: how GeoUp has grown since the first version
With the first version in use, development continued. According to the announcement on geoup.io, the second version grew out of user feedback. Three new developments are at its center:
- Features for agencies: Agencies manage several client brands from one login and invite their team with editor access. Dashboard and reports can carry the agency's own logo (white label). With the free quick check, which delivers a first analysis based on sample questions, agencies can see during the very first meeting whether AI systems mention a prospect.
- Reports: Results are available as PDF reports. Quick check results can also be shared as a web report via link, which colleagues and management can open without logging in.
- New design: The interface was redesigned from the ground up: clearer, more intuitive and designed so that every next step feels natural.
The measurement has grown too. Today GeoUp queries five AI systems: ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. The full analysis asks 30 questions per AI system and runs regularly, up to weekly. Market Battle Mode sets a business's visibility side by side with that of its competitors and shows which sources the AI cites for them.
The results lead to concrete measures. Schema & Snippets generates structured data and quotable text snippets and adds them to the website with one click. Content Automation writes content in the brand's voice, the AI directory listing automatically creates a structured profile for every business in the GeoUp directory, and Authority Building finds the portals that AI systems rely on and checks whether name, address and phone number match everywhere.
Another new addition is an MCP server. The Model Context Protocol (MCP) is an open standard through which AI applications such as Claude or ChatGPT connect to external systems. This lets customers query their GeoUp data directly in their AI assistant, while AI agents can read a business's public brand profile. Our article MCP for software providers explains how this works for software providers in general.
What you can take away for your own project
A clear question sets the direction. For GeoUp, it was: Do AI systems recommend a business or not? From that, it followed what the first version needed to do: measure, and offer tools for improvement.
The first version is complete without being able to do everything yet. With GeoUp, the path from measurement to action worked, along with billing and administration. Another AI system, agency features and reports came afterwards, based on what we learned from real use.
Experience and AI complement each other. AI brings speed to the implementation, and experienced developers make the decisions that matter: architecture, data model and quality.
Frequently asked questions
How long does it take to develop a SaaS product?
That depends mainly on the scope. Two senior developers built the first version of GeoUp in two months, including billing, admin dashboards and automation. You can get a free estimate for your project with our AI project estimator.
What belongs in the first version of a SaaS product?
The core workflow that solves your customers' problem, plus everything you need to sell and operate it: user accounts, billing and an admin area for your team. Plan further features as later stages and prioritize them based on feedback from your first customers.
Does AI write the code at njudev?
AI handles the routine work for us. We bring 20 years of experience to the decisions, details and delivery. For GeoUp, architecture, logic and data model were settled before the AI began, and every result was reviewed and approved by senior developers.
Why does an agency build its own products?
We build for others and for ourselves, because you understand what a client needs best when you are an entrepreneur yourself. Besides GeoUp, the AI shopping agent byclaw.io is also one of our own products. What we learn along the way flows into every client project.
Conclusion
GeoUp shows what a clear question, experienced developers and AI can create in a short time. Two senior developers built the first version in two months, with measurement, tools for improvement, billing, administration and automation. Since then, the product has grown with feedback from its users. With the same way of working, your idea can also become a platform that launches early and grows with your business.
Planning your own SaaS product?
Describe your idea in our AI project estimator. You will receive a free estimate by email, and we will get back to you within 24 hours.
Learn more about our services for platforms and marketplaces and SaaS products. Or simply write to us at our email address.
With the GeoUp quick check (in German), you can see within moments, based on sample questions, whether ChatGPT, Perplexity, Claude, Gemini and Google's AI Overviews mention your business.
Sources
- njudev: case study "GeoUp.io, SaaS Analytics Platform". https://njudev.com/en/work/geoup
- njudev: homepage, sections "How we work" and "Cases". https://njudev.com/en
- GeoUp: "KI-Sichtbarkeit für Unternehmen" (AI visibility for businesses), homepage, in German. https://geoup.io
- GeoUp: "Das neue GeoUp ist da: tiefere Analysen, teilbare Reports und nahtlose Anbindung an Ihre Systeme" (the new GeoUp is here), July 5, 2026, in German. https://geoup.io/news/geoup-neue-version-2026
- GeoUp: "KI-Sichtbarkeit für Agenturen unter eigener Marke" (AI visibility for agencies under their own brand), in German. https://geoup.io/fuer-agenturen
- GeoUp: "Market Battle Mode", in German. https://geoup.io/market-battle
- GeoUp: "Schema & Snippets", in German. https://geoup.io/schema-snippets
- GeoUp: "MCP-Server: Ihre KI-Sichtbarkeit. Direkt im KI-Assistenten." (MCP server), in German. https://geoup.io/mcp
- GeoUp: "KI-Sichtbarkeitsanalyse" (free quick check), in German. https://geoup.io/ki-sichtbarkeitsanalyse
- Model Context Protocol: "What is the Model Context Protocol (MCP)?". https://modelcontextprotocol.io/docs/getting-started/intro
- Schema.org: "About". https://schema.org/docs/about.html
- Stripe: "Stripe Billing: Wiederkehrende Zahlungen und Abonnements" (recurring payments and subscriptions, Germany). https://stripe.com/de/billing