Docker Launched Cloud Sandboxes for AI Agents
The new platform enables AI agents to run tasks in cloud-based microVMs after a developer's local machine is offline.
Updated on Sept. 24, 2026 in Artificial Intelligence
Live Poll
Do you trust AI agents to handle complex, high-stakes tasks without human supervision?
Docker has launched Docker Cloud Sandboxes, a managed service designed to execute AI agent workloads in the cloud. The platform utilizes microVMs to provide isolated environments that persist after a user's laptop is powered down.
Why it matters
Modern agentic workflows often require high-stakes, long-running processes that are impractical to maintain on local hardware. This launch addresses the demand for scalable agent execution that balances performance with governance and security controls.
The platform supports scaling from 1 to 16 vCPUs per task and achieves boot times in the range of hundreds of milliseconds. It leverages Kits, an open specification that packages agents, their tools, and access rules into a single OCI image artifact.
The players
Docker
A software company focused on containerization tools and cloud-native development platforms.
Cloud Native Computing Foundation
A Linux Foundation project that maintains and integrates open-source cloud-native infrastructure technologies.
The details
Docker Cloud Sandboxes provide microVM isolation—a lightweight virtual machine technology that offers faster startup times and smaller footprints than traditional VMs—to contain agentic workloads. By pointing agents to Docker-managed infrastructure, developers offload task execution to the cloud while maintaining security through access rules embedded directly within the OCI (Open Container Initiative) image. This architecture allows for unattended, long-duration processing while ensuring that policies are strictly enforced.
Timeline
Docker introduced local sandboxes earlier in 2026.
Docker launched Docker Cloud Sandboxes and Kits on September 24, 2026.
The Tech Race
This development marks a significant move to standardize how AI agents are packaged and deployed across different environments. By utilizing the OCI image standard, Docker aims to ensure that complex agentic workflows remain portable and governed in a way that proprietary runtimes often fail to achieve.
Developers can now offload resource-intensive agent tasks to managed cloud environments, freeing up local computing resources. Those looking to build compatible workflows should monitor the upcoming submission of the Kits specification to the Cloud Native Computing Foundation.
The takeaway
This release shifts the burden of agentic execution from local machines to the cloud, allowing for reliable long-running AI processes. Users should monitor the public repository for the Kits specification to ensure their agentic tooling remains compatible with emerging industry standards.
Further reading
For more context on how developers are automating high-stakes tasks, see our Artificial Intelligence section.
More information
To review the full product specifications and technical documentation, visit the Docker Cloud Sandboxes product page.
Live Poll
Do you trust AI agents to handle complex, high-stakes tasks without human supervision?
