Show HN: Token Beggar, route coding agent subtasks to cheaper models (github.com)

Token Beggar is a new plugin for Claude Code designed to optimize resource usage by routing specific coding tasks to cheaper models before assigning them to more powerful ones. This innovation allows for small tasks to remain within the chat while larger, more complex tasks—such as debugging—are sent to less expensive workers for execution. In trials, Token Beggar demonstrated a cost reduction of 29% on average for 18 real bug fixes, confirming its efficacy by achieving similar success rates in issue resolution (69% compared to 70% when using the model alone). The significance of Token Beggar lies in its potential to enhance efficiency in AI/ML workflows, reducing operational costs while maintaining performance. Technically, it employs a clear workflow that identifies task complexity and routes it accordingly, operating entirely on the user's machine without sending any personal data externally. For developers and teams who engage with large-scale models like Opus 5.5, Token Beggar offers a practical approach to cost management and task optimization, ensuring that they can leverage advanced AI capabilities without incurring excessive expenses.