Back to Kchime

Dev Log

Consumer

Building Kchime

Reply naturally in English, every time.

KChime — AI reply assistant with smart replies, practice scenarios, and daily streaks
Live

Architecture

KChime runs as both a web app and an iOS keyboard extension. The reply engine accepts an incoming message, analyzes tone and context, then generates natural reply suggestions. A practice mode creates real conversation scenarios to build English fluency. Daily streaks and progress tracking keep users engaged. The iOS keyboard extension communicates with the backend via a shared App Group container.

Tech Stack

Next.jsSwift (iOS)OpenAISupabase AuthStripeTailwind CSS

Try KChime

kchime.com/

Build Timeline

Core reply engine with 4 tone modes

Built the reply suggestion engine — paste any incoming message and get natural replies in 4 tones: casual, professional, friendly, and direct. Each reply includes a confidence score. Added "Fix My Message" mode that polishes user-written drafts to sound more natural while preserving their intent.

LLMTone AnalysisReply Engine

Conversation practice with 50+ scenarios

Created a practice system with 50+ conversation scenarios across categories: workplace, social, interviews, small talk, conflict resolution, and dating. Each scenario includes an AI conversation partner that responds naturally and provides feedback on the user's replies. Added voice mode using the Web Speech API.

Practice ModeAI PartnersVoice

iOS keyboard extension + daily learning

Shipped the iOS keyboard extension in Swift — users can get reply suggestions directly from any messaging app without switching contexts. Added a daily learning feed with 3 new phrases, idioms, and cultural context tips. Integrated Supabase Auth for cross-platform sync and Stripe for Pro subscriptions.

iOSSwiftKeyboard Extension

More Dev Logs