Stop Rationing Tokens: Let the Harness Pick the Model — Kimchi by Cast AI

The Kimchi coding platform addresses the rising costs of LLM token usage by implementing an autonomous harness that dynamically selects the most cost-effective model for specific tasks without compromising quality. By utilizing an automated, multi-model approach, the platform achieves significant savings, with internal data showing a 2.5-fold reduction in costs despite a 1.5-fold increase in token volume. Key features include "Feynman," which manages long-running, multi-step coding tasks with human-in-the-loop milestones, and "Teleport," a remote sandbox solution that allows developers to maintain persistent coding environments across devices and network interruptions. Additionally, "Kimchi Studio" facilitates team collaboration by providing a visual interface for tracking, reviewing, and managing agentic workflows. This open-source framework shifts the focus from individual prompt engineering to scalable, managed software development lifecycles that prioritize efficiency and team-wide visibility.