Documentation
Why it exists, how to use it, and how it is built: a free, open-source AI coach for the new 2026 TOEFL, powered by Meta's Muse models.

Read in this order
Start with why the app exists, see it in action, then go as deep into the technology as you like.
Start here
Old books, high prices and no feedback in Dili, Timor-Leste, and how that became a free AI coach, now also in VR on Meta Quest.
Using the app
A tour of the app with screenshots: study plan, all 12 practice types, vocabulary, progress, mock test, certificates and the VR Study Room.
Using the app
Why the TOEFL is better practised in VR, and how to use the hands-first study room on Meta Quest: getting in, gestures, the 9 tasks, the 3D speakers and interviewer, and the progress wall.
Technical
Architecture diagrams: Vercel, Render, PostgreSQL, AWS S3, every Meta model, the job queue, mock-test scoring, the database, the VR Study Room and signing in on Quest.
Technical
What one study plan really costs in AI calls: measured tokens, pictures and voice from a real plan, priced with Meta's and ElevenLabs' published rates.
Technical
The project README: features, models and architecture, then how to set up, run and deploy your own copy.
User guide
The user guide walks through the real app with screenshots: the AI study plan, all 12 practice types, vocabulary from your own questions, progress, the timed mock test and verifiable certificates.
Open the user guideHow it works
A Next.js web app on Vercel, a background worker and PostgreSQL on Render, and private media in AWS S3. The worker and the web app share work through a job queue in the database.
muse-spark-1.3Questions, vocabulary, study schedules, AI scores and coachingmuse-spark-1.2Automatic fallback for textmuse-image-1.0Pictures for practice questionsmuse-voice-transcribe-1.0Speech-to-text for spoken answerseleven_multilingual_v2Voices in five English accents
System architecture and 4 more diagramsStart free with a personal AI study plan and a full mock test, or deploy your own copy from GitHub.