Fluent transforms Claude Code into a personalized AI language tutor, offering adaptive practice powered by cognitive science principles like spaced repetition and active recall. It targets users seeking focused, efficient language mastery without distractions, providing intelligent tracking and a private learning environment. The system adapts to individual learning styles and goals, making language acquisition more effective and enjoyable.

How It Works

This system leverages Claude Code as an AI tutor, implementing evidence-based learning methodologies. Its core approach includes Spaced Repetition (SM-2 algorithm) for optimal review intervals, Active Recall to strengthen memory retrieval, and Adaptive Difficulty that adjusts challenges to maintain a 60-70% success rate. Learner data, including progress, mistakes, and mastery levels, is stored locally in six JSON databases, ensuring privacy and enabling the AI to continuously learn about and adapt to the user's specific needs and weaknesses.

Quick Start & Requirements

• Primary install / run command:
  • Plugin: claude plugin marketplace add m98/fluent && claude plugin install fluent@m98 followed by /fluent-setup and /fluent-learn.
  • Clone: git clone https://github.com/m98/fluent.git, cd fluent, then claude and /fluent-setup.
• Non-default prerequisites and dependencies: Claude Code, Python 3.8+ (standard library only), Bash (for PreCompact hook; WSL or Git Bash recommended on Windows).
• Estimated setup time or resource footprint: Minimal setup time; low resource footprint as it relies on standard libraries and local data storage.
• Links: Claude Code Documentation (implied), SM-2 Algorithm, CEFR Levels

Highlighted Details

• Multi-Language Support: The system is language-agnostic, allowing users to learn any language.
• Evidence-Based Methods: Integrates Active Recall, Spaced Repetition (SM-2), Immediate Feedback, Interleaving, Comprehensible Input (i+1), and Desirable Difficulty.
• Adaptive Difficulty: Dynamically adjusts exercise difficulty based on user performance to optimize the learning curve.
• Privacy-Focused: All learner data remains local, is automatically excluded from Git, and involves no external tracking or telemetry.
• Comprehensive Tracking: Utilizes six JSON databases to meticulously track error patterns, mastery levels, review schedules, and overall progress statistics.

Maintenance & Community

Contributions are welcomed via pull requests, with priority areas including language-specific optimizations, audio/visual features, performance enhancements, and accessibility. Support for bugs and questions is available through GitHub Issues.

Licensing & Compatibility

This project is licensed under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

The system is strictly dependent on Claude Code. Certain automated hooks (PreCompact) require Bash, necessitating WSL or Git Bash on Windows. While Python hooks function on native Windows Python, the full hook system may require a Bash-like environment for complete functionality.