Cline Bot Inc. is expanding its open-source AI coding harness beyond software development into general-purpose tasks. The company aims to provide users with the flexibility to select cost-effective inference models for specific workloads, rather than relying solely on frontier models.
The company has spent two years developing an evaluation system based on real-world traffic to optimize its harness for both large and small models. Cline recently open-sourced these evaluations for open-weight agents to support modular architecture and self-correction loops.
Usage of the platform has surged, with token consumption increasing 20x over the last four months. This growth accelerated in May following the release of several open-weight models. Much of this inference activity is processed on the CoreWeave serverless platform.
Cline is extending its reach beyond integrated development environments and command-line tools with the launch of Cline Desktop. This open-source application is designed to facilitate the use of open-weight models for general knowledge work and non-technical users.
