Thu, 16 Jul 2026 — Cast AI launched general availability of Kimchi Coding, an autonomous AI coding agent for software development teams, moving the product out of early access into production use. Akamai is already using the system.
Kimchi Coding uses a multi-model orchestration approach: routes work to different models based on complexity and cost rather than relying on a single large model for every task. Cast AI frames this as matching enterprise demand to reserve frontier models for hard tasks and open-weight models for routine work. Deployment options include customer VPC or dedicated Nvidia B300 GPUs managed through Cast AI infrastructure.
Company-claimed shadow-mode evaluations vs commercial-model-only baseline: 2.5× cheaper while matching or exceeding quality on specification matching and test pass rates. Features include spending controls (API keys to org budgets), automated shutdown of runaway agent loops, and dashboards tracking cost by developer, team, and project. Governance: hard spending caps, data sovereignty, self-hosted inference; ISO 27001, SOC 2 Type II, GDPR-compliant per company.
Quotes: Laurent Gil (president/co-founder): “General availability means that enterprises are now using Kimchi Coding at scale… The economics of AI coding are broken when you rely on a single commercial model.” Dekel Shavit (senior director of engineering, Akamai): multi-model approach keeps frontier quality while cutting token costs; governance suite praised for data sovereignty, cost observability, and budget control.
Cast AI background: Kubernetes and GPU optimization unicorn (Pacific Alliance Ventures / Shinsegae); customers include BMW, Cisco, FICO, Hugging Face, Swisscom. Kimchi extends the business into AI software development tooling.