Kimi Launches Kimi Hosted Agents for Managed AI Agent Deployment
Kimi has launched Kimi Hosted Agents, providing a Harness-as-a-Service cloud environment that offers 24/7 managed execution for long-running tasks and multi-agent collaboration. The platform includes over 140 skills, 30 plug-ins, independent sandboxing, and initial templates tailored for financial use cases. This release significantly lowers the barrier for enterprises to deploy production-grade AI agents by offloading infrastructure maintenance, security sandboxing, and tool integration to a managed cloud service. It enables businesses to quickly transition from prototyping internal and customer-facing workflows to reliable, scalable operations. The service is priced at ¥0.072 per hour for sandbox runtime, alongside model inference costs and plugin fees. Additionally, it features Kimi Inference Hook, an independent security component allowing enterprises to filter inputs and tool outputs prior to model processing.
## BACKGROUND
An AI agent harness, or runtime scaffolding, is the software infrastructure surrounding a large language model that drives model and tool calls, manages conversation state, and keeps the agent progressing through multistep tasks. Sandboxing in this context provides a secure, isolated execution environment to prevent unauthorized system access or high-risk actions during agent execution.