Moonshot AI has launched Kimi Hosted Agents, a managed cloud service designed to package the agent runtime, or harness, as a product. According to IT Home and the company's open-platform documentation, this service removes the need for enterprises to build and maintain that layer themselves.
The offering, framed as Harness as a Service, allows teams to configure agents on Kimi's infrastructure to receive a 24/7 hosted environment. This environment is intended for long-running tasks and multi-agent collaboration. Companies can point these agents at internal systems for research, monitoring, and operations work, or embed them in customer-facing products to serve as support or research assistants.
Kimi Hosted Agents provides isolated sandboxes with network-access controls and session boundaries that operators can stop at any time. The service also includes execution logs that can be exported for audit. IT Home reports that usage reports break out sandbox hours, model inference, and plugin calls, with a sandbox rate of 0.072 yuan per hour plus additional charges for models and plugins.
The platform currently lists more than 140 skills and more than 30 plugins. These cover content generation, office documents, professional data queries, and web search. Developers have the option to attach their own tools, data sources, and MCP services.
For security, an optional component called Kimi Inference Hook allows an enterprise's own or third-party safety service to inspect task inputs and tool returns before each model call. This ensures only content passing the company's rules proceeds to inference, though the hook is currently limited to enterprise-verified customers. Official documentation also describes self-hosted environments where sandbox tools run on customer-controlled machines while orchestration remains on the platform.
Alongside the managed runtime, Moonshot is shipping a first set of official agent templates. IT Home highlights eight finance templates—including IPO tracking, morning briefs, portfolio monitoring, modeling, and deep research—while the documentation lists a broader first wave of 12 templates that also include SRE incident response and game operations. These templates can be copied into a project space and connected to internal reports or private databases.
The launch expands Kimi's product line from models and desktop agents into a hosted runtime that Moonshot says it will maintain and upgrade. This comes a day after Huawei Ascend announced day-zero inference support for the open-weight Kimi K3.1 Pro model, underscoring a wider push to make Kimi agents easier to run in production environments.
