A behind-the-scenes case study of Ecomtoro, our AI product research SaaS for Amazon, Shopify and TikTok Shop sellers: the problem, product decisions, AI agents, stack, billing, MCP and lessons learned.
In short
A behind-the-scenes case study of Ecomtoro, our AI product research SaaS for Amazon, Shopify and TikTok Shop sellers: the problem, product decisions, AI agents, stack, billing, MCP and lessons learned.
Ecomtoro is an AI product research tool for online sellers. A seller describes a product idea, and AI research agents check real sales history, search demand and competition across Amazon, Shopify and TikTok Shop markets in the US, UK and Germany, then work out the margin after every fee. We designed and built it end to end: the product, the AI agents, the dashboard, billing and the landing page.
This case study shows the decisions behind it. If you are planning your own SaaS or an AI-powered web app, these are the same questions you will face.
Sellers lose money on products that look good in a spreadsheet and fail in reality. The usual research process is manual and scattered: browse marketplace listings, guess demand from a few reviews, look up search trends, then estimate margins in a separate sheet that forgets half the fees.
We wanted one place where a seller could ask "Should I sell this?" and get an answer based on data, with the profit math done properly. The product had to be fast enough to test many ideas, and honest enough that sellers could trust it before buying stock.
Many research tools show dashboards full of numbers and leave the conclusion to the user. We designed Ecomtoro around a conversation: describe the idea, and the agents come back with a structured verdict, the evidence behind it and the numbers that matter. The dashboard exists to explore the evidence, not to replace the answer.
A product with strong demand can still lose money once marketplace fees, fulfilment, shipping and advertising are counted. The profit calculator takes every fee into account per marketplace, so the margin a seller sees is the margin they would actually get.
Product research is not a one-off. A watchlist with alerts lets sellers follow promising products and react when demand or competition changes. This turns a one-time check into a habit, which matters for a subscription business.
The core of Ecomtoro is a research agent that works through a set of focused tools. Each tool has one clear job:
| Tool | What it does |
|---|---|
| Product details | Pins down exactly which product and market the seller means |
| Market data | Pulls sales history and search demand over time |
| Competition | Checks how crowded the market is and how strong the leaders are |
| Economics | Runs the margin with every relevant marketplace fee |
| Decision | Combines the evidence into a clear verdict with reasons |
Splitting the work into focused tools made each part easier to test and improve. It also made the output easier to explain: every verdict comes with the evidence behind it.
We built Ecomtoro on the same stack we use for client web applications:
Moving long research runs into background jobs was one of the most important decisions. The user gets immediate feedback and progress updates, and a slow external source never blocks the interface. We explain these choices in choosing a backend for your SaaS.
Sellers increasingly work inside AI assistants. So we built an MCP connector: sellers can ask Ecomtoro questions directly from Claude and ChatGPT, and the assistant calls Ecomtoro's research tools behind the scenes. The same business logic serves the web app and the connector, so results are consistent wherever the question is asked.
This is a pattern we now recommend to many SaaS products: if your users already spend their day in an AI assistant, meet them there.
Subscriptions sound simple until you handle trials, upgrades, downgrades, failed payments and usage limits. Stripe handles payments and invoices; our backend keeps the source of truth for what each account may do, based on webhook events. Every limit is enforced on the server, never only in the interface.
The landing page has one job: make a seller understand the promise in seconds and try it. The headline speaks the seller's language ("Find what to sell. Know it pays before you buy."), the page shows the real product interface instead of abstract graphics, and the call to action is repeated after each section. You can see it scrolling in our portfolio.
The same building blocks fit many businesses: a clear core workflow, AI agents with focused jobs, a solid backend with background processing, subscriptions, and a fast landing page. Whether you are building a SaaS or adding AI to an internal tool, we can help with web applications, AI automations and product design. Planning your first version? Read MVP in 6–8 weeks, then tell us about your idea.
An AI product research SaaS for Amazon, Shopify and TikTok Shop sellers. It checks demand, competition and margins before a seller buys stock.
No, Ecomtoro is our own product. We use the same process and stack for client projects.
MCP (Model Context Protocol) is a standard way for AI assistants such as Claude and ChatGPT to use external tools. Ecomtoro's connector lets the assistant run Ecomtoro research on the user's behalf.
Yes. We design and build SaaS products and AI-powered web apps from discovery to launch, including agents, billing and the marketing site.
