South African startup Exten AI turns plain-language descriptions into working software — and its founder argues the hard part is not the code, but everything underneath it.
Launched this year by Luncedo Simelane, the platform provisions a full backend alongside each generated application: an isolated database, file storage, secure user authentication and the runtimes needed to host autonomous AI agents.
“It is basically a digital tech co-founder to a non-technical entrepreneur with a great software idea, delivering a production-ready stack in under 30 minutes,” Simelane told Disrupt Africa.
The plumbing problem
Simelane taught himself to code before leaving his LLB studies to focus on software architecture full-time. His diagnosis is that infrastructure gaps, not ideas, are the constraint.
“Global text-to-code platforms primarily serve existing developers to make them code faster, or stop at creating superficial visual frontends,” he said. “Solo builders and small business merchants hit a brick wall because they don’t know how to deploy backend systems, structure relational databases, or handle security authentication.”
On positioning: “We sit between global frontend prompt tools, which skip the database entirely, and complex enterprise low-code systems like Microsoft Power Apps, which are far too expensive and rigid for emerging market creators.”
That is a narrower claim than most startups make, and a checkable one.
Where it actually is
The first public release drew over 100 signups. Demand outran the servers immediately, and Exten AI moved into closed-cohort testing to keep infrastructure stable.
Three apps built by early users are live: LedgerAI for accounting automation, EduAgent for teachers’ lesson planning, and LuminaAI, a custom communication dashboard.
A pilot began at the end of September with the University of Fort Hare’s Centre of Entrepreneurship and Rapid Incubator, where student founders will build MVPs on the platform. An early proof of concept was endorsed on Sunday Sessions by Professor Mamokgethi Phakeng, former vice-chancellor of the University of Cape Town.
Scaling was the problem
“Having dozens of users simultaneously asking an AI engine to map out, provision, and launch complex database schemas live on the cloud caused massive infrastructure queues early on,” Simelane said. “It required weeks of deep-dive architecture rewrites to isolate and stream these automated builds.”
That is a specific technical account of a specific failure, which is rarer in startup coverage than it should be.
The money
Exten AI is fully bootstrapped and raising R1.5 million ($90,000) to scale its server layers and support thousands of concurrent builds.
It runs a tiered SaaS subscription. Generating and testing apps is free; users pay monthly once an app goes live publicly, with pricing scaling by server load, database storage and AI agent usage.
“Because we’ve prioritised architecture optimisation during our closed-cohort phase to perfect our user unit economics, early revenues have been minimal,” Simelane said.
No pricing tiers, user numbers beyond the signup figure, or revenue figures were disclosed, and the company has not said which models power the generation layer.
The market he’s aiming at
Exten AI is focused on South African SMEs and solo founders, with plans to expand to tech hubs and university incubators across sub-Saharan Africa.
“Emerging markets are experiencing a severe software developer shortage alongside an explosion of young entrepreneurs who need localised digital tooling to survive and scale,” Simelane said.
Based on reporting by Disrupt Africa.
