Four different people on the engineX team, working on four unrelated client projects, ran into the exact same wall. Product lead Ohene-Addo-Agyekum tried building a barbering shop website on Lovable and could only wire up Stripe. Frontend engineer Essilfie Prince Bondzie hit the identical limit on an e-commerce build. AI engineer Elijah Darkeh Agbedam tried Lovable, then v0, then bolt.dev, and got the same answer three times. Fullstack engineer Justice Osei Sarfo ran into it building a booking site for a local spa.
None of them knew, at the time, that the others had already found the same dead end. "Most companies do not have their focus on Africa," is how the team eventually put it, once they compared notes, and the gap became the founding premise of engineX.
The Gap Nobody Built For
Business software in Ghana is neither cheap nor fast to stand up, often taking a month or more before a company can even begin testing. The newer wave of AI website builders promised to shortcut that, and inherited the exact same blind spot. They're optimized, by default, for a credit-card-first world. "Global platforms are built for that world," Sarfo says, "leaving a massive infrastructure gap that requires local developers to do the heavy lifting of bridging advanced AI with African payment realities."
What EngineX Actually Does
A user describes their business in plain language, "a booking site for my hair salon in East Legon," and a coordinated team of AI agents takes over: building the frontend, the backend, and a working payment checkout, with the user supervising and approving along the way. Where payment integration is needed, a lead agent asks directly for the relevant details. Within minutes, a business owner can share a live link where customers pay by MTN Mobile Money, Telecel Cash, or AT Money, with funds landing in the owner's own wallet.
The hardest problem underneath that simplicity wasn't the interface, it was stopping the AI from inventing things where money was involved. Sarfo describes building "a schema translation and deterministic generation layer" that forces the model to produce structured, accurate payment payloads and correctly mapped webhook callbacks without hallucinating an endpoint or a payload key. Getting an AI to write ordinary frontend code, he says, is comparatively easy. Getting it to reliably handle money movement is a different order of problem entirely.
The team's first instinct was to let a single agent handle everything end to end, including writing custom payment logic from scratch for each new site. It didn't hold. A single agent handling every task reduced output quality, and letting a model freelance on financial logic produced invented endpoints and misconfigured payloads. The team rebuilt around a stricter architecture: multiple specialized agents working in coordination, each interacting with standardized, pre-built infrastructure blocks instead of writing financial code from scratch.
The Moolre Layer
EngineX's integration started with Collections, used both to power payment collection inside every system the platform builds and to collect engineX's own user subscriptions. Later in the competition, the team added SMS setup for the businesses it builds for, expected to ship alongside disbursement capabilities. The team has also built its own MCP (Model Context Protocol) server, letting its AI agents work with Moolre's infrastructure directly, with an eye toward eventually extending the same agent-facing layer to other payment, email, and SMS providers across the continent.
The clearest snag wasn't technical so much as procedural. After submitting verification documents through Moolre's dashboard, the team's account was marked "Verified," and they assumed that meant they were cleared to build. It turned out "Verified" and "Activated" were separate states the dashboard didn't clearly distinguish, and the team had to contact support directly before development could continue. Once resolved, the rest of the integration went smoothly.
What's Next
Team members describe a recurring, almost validating reaction from users: disbelief that engineX is an African-built company, once serious enough that a user tried to hack the system just to understand how it worked. Sarfo's broader read on building here isn't about talent or code. "The hardest part is dealing with the unmapped fragmentation of real-world infrastructure," he says. "You aren't just writing software; you are actively pioneering bridge layers over broken systems." Bondzie puts the same idea more plainly: copying the Silicon Valley playbook wholesale doesn't work, and founders have to blend it with what they actually know about how business happens locally.
EngineX pitches live in the final stage of the Moolre Startup Cup, with winners announced at the Grand Finale on August 29, 2026.
