Kimchi is a governed AI coding platform that brings coding agents, model inference, and spending controls together. Built around open-source models, it helps engineering teams use AI to plan, build, and review code while managing costs and protecting data. Its enforcement layer applies spending, data, deployment, and model-routing policies before requests reach a model. Teams can set hard budget limits by user, team, API key, or model, and monitor usage and spending attribution in real time.

The platform includes Kimchi Coding, a command-line coding agent; a serverless inference API for open-source models; Studio, a visual workspace for AI agents; Teleport, for handing sessions off to the cloud; and AI-powered pull-request review. Organizations can use the serverless service or deploy on Kubernetes in AWS, GCP, Azure, or on-premises environments. Kimchi states that prompts, code, and outputs are excluded from model training, and that deployment options can keep model data within an organization’s environment. Developers can work through coding tools their teams already use, while security and platform teams manage policies centrally.