Show HN: Triplecheck – Review your code free with local LLMs (github.com)

Triplecheck, a new open-source tool for automated code reviews, has been introduced to the AI/ML community. It leverages local large language models (LLMs) to provide thorough, iterative code evaluations without incurring API costs. With its unique three-agent system—Reviewer, Coder, and Judge—Triplecheck iteratively finds bugs, proposes fixes, and scores code quality in a multi-pass review cycle. This approach ensures deeper insights than conventional linting tools, allowing developers to receive actionable code corrections that enhance the functionality and reliability of their software. The significance of Triplecheck lies in its ability to operate entirely locally, thus offering unlimited, cost-free services to developers. It supports multiple programming languages and enables integration with various LLM backends, including models like Qwen and Claude. Key technical features include the ability to automatically split large repositories into manageable segments, execute real fixes instead of mere suggestions, and employ a voting mechanism to filter out noise from the findings. With planned GitHub integration and support for IDE extensions, Triplecheck is positioned to become a vital asset for developers seeking to enhance their coding practices through advanced AI capabilities.