AccessBuild, a new AI-powered accessibility audit tool, has officially shipped its backend and public dashboard after a four-day build sprint. The project, created by developer Okeke Chukwudubem, addresses a critical infrastructure gap: the lack of accessibility labels in African mobile applications. By scanning Android UI trees and assigning A-to-F scores, the tool provides quantitative data on app usability for visually impaired users and AI agents alike.

The Discovery That Sparked the Build

The project began when Chukwudubem noticed his Phone Agent could not interact with local banking apps. This was not due to agent errors, but because the apps lacked content descriptions and screen reader support. Initial testing of 30 apps yielded stark results: 100% of Nigerian banking and government apps scored an 'F' (below 25% labeled), while global giants like WhatsApp and Google Maps achieved 'A' ratings (90%+ labeled). This disparity highlighted that accessibility neglect locks out both human users with disabilities and automated AI agents.

Technical Stack and Implementation

The architecture is lean and practical, prioritizing speed and low cost. The backend runs on FastAPI with Python, deployed on Render’s free tier, while data persistence is handled by Supabase (PostgreSQL). The AI component leverages Google Gemini to parse Android UI Automator XML dumps, generating scores and recommendations in under two seconds per audit. The pipeline is fully automated: the Phone Agent dumps the UI tree, AccessBuild parses and scores it, and the results are saved to the database for public display.

Roadmap to Scale and Compliance

With the core engine live, the project is entering Phase 1 of its roadmap: scanning over 100 Nigerian apps across banking, telecom, health, and education to build the first open accessibility database for Africa. Future phases focus on awareness, developer tooling, and regulatory compliance. The team aims to partner with disability advocacy groups and regulators to make accessibility scores mandatory, eventually offering certification for compliant apps and expanding to Ghana, Kenya, and South Africa.

Key Takeaways

  • AccessBuild uses a FastAPI and Supabase stack to deliver sub-2-second audit responses.
  • Testing revealed that 100% of Nigerian banking and government apps fail accessibility standards.
  • The project treats accessibility as infrastructure, arguing that unlabeled apps block both AI agents and visually impaired users.
  • The roadmap includes creating a self-serve developer portal and seeking grant funding to move from prototype to scale.

The Bottom Line

Accessibility is infrastructure, not a feature. AccessBuild proves that automated auditing can expose systemic neglect in African apps, but real change requires regulatory pressure and funding to move from prototype to standard.