Docker Cloud Sandboxes Enable Seamless Transfer of Long‑Running Coding Agents Between Laptop and Cloud

Docker announced the launch of Cloud Sandboxes, an extension of its existing Docker Sandboxes product line. The core promise is the ability to shift a coding agent from a developer’s laptop to a Docker‑managed cloud instance with a single command, eliminating the need for manual reconfiguration or data migration.

The underlying isolation model remains identical to the local version. Each Cloud Sandbox runs inside a micro‑VM that provides its own kernel, Docker daemon, secret store, and network policy. This one‑to‑one correspondence ensures that workloads behave the same way whether they are executed on a laptop or in the cloud.

Design for Long‑Running Tasks

Docker positions Cloud Sandboxes as a solution for workloads that run for several hours. Because the cloud instances are always on, developers no longer depend on the availability of their laptop for the duration of a task. The continuous‑up environment is billed on a pay‑as‑you‑go basis, allowing teams to spin up resources only for the time they need them.

The pricing model is transparent and granular. Docker charges per second, with the smallest instance type, called Micro, priced at $0.07 per hour, and the largest, XL, at $1.12 per hour. Data egress, volume usage, and public image hosting are offered at no additional cost, simplifying budgeting for extended runs.

Pre‑Configured Agent Kits and Custom Extensions

To accelerate adoption, Docker supplies pre‑configured kits for a range of major coding agents, including Claude Code, Codex, Copilot, Antigravity, Open Code, and Hermes. These kits contain all necessary dependencies and configuration files, allowing a developer to start a sandbox with an agent in seconds.

Beyond the supplied kits, Docker permits the addition of custom kits. Organizations can package proprietary tools or specialized libraries into a kit and load it into a Cloud Sandbox, preserving the same isolation guarantees as the built‑in options.

Overcoming Local Sandbox Limitations

According to the security‑focused analysis firm Publickey, two primary constraints of local Docker Sandboxes are resolved by the cloud extension. First, local sandboxes require the laptop to remain powered on for the entire execution of long‑running tasks. Second, a single laptop cannot efficiently host more than a hundred sub‑agents concurrently, limiting parallel experimentation.

By moving the sandbox to a cloud host, developers can keep their laptops free for other work while the sandbox continues uninterrupted. The cloud’s scalable resources also enable many more sub‑agents to run in parallel, expanding the throughput of AI‑assisted coding pipelines.

Seamless Migration with sbx move

Both Docker’s announcement and the Publickey report describe the migration command, sbx move, as a key feature. When invoked, sbx move captures the current file system state of the active sandbox, packages it, and reconstructs the environment on the target side—whether that is the cloud or the laptop. The process is designed to be continuous, meaning the coding agent experiences no interruption during the handoff.

The migration includes all mounted volumes, environment variables, and secret references, preserving the exact runtime context. Because the underlying micro‑VM architecture is consistent across both locations, the reconstructed sandbox boots with the same kernel version and Docker daemon configuration, further reducing the risk of incompatibility.

Pricing and Cost Considerations

While the per‑hour rates appear modest, Docker’s pay‑as‑you‑go model charges by the second. Organizations must therefore monitor usage to avoid unexpected spend, especially when scaling to larger XL instances for intensive workloads. The free egress and volume allowances help contain costs, but network traffic beyond the free tier could still impact the total bill.

• Micro instance – $0.07 per hour
• Small instance – intermediate price (not disclosed)
• Large instance – intermediate price (not disclosed)
• XL instance – $1.12 per hour

Docker does not publish a detailed breakdown for the intermediate instance sizes, leaving enterprises to request a quote or test the offering to determine the optimal size for their workloads. The lack of granular pricing for those tiers introduces an element of uncertainty when planning large‑scale deployments.

Future Governance Features

Docker has hinted at a forthcoming Docker AI Governance feature aimed at providing centralized policy controls for enterprises. However, the company has not supplied a release date or detailed capability list. As a result, organizations cannot yet rely on built‑in governance tools to enforce corporate AI usage policies within Cloud Sandboxes.

The absence of concrete governance details means that teams must currently implement external controls—such as network firewalls, access‑management solutions, and audit logging—to meet compliance requirements. Until Docker releases the promised governance suite, the responsibility for policy enforcement remains with the customer.

Practical implications for organisations are clear. Cloud Sandboxes remove the need for developers to keep laptops powered for extended AI‑driven coding sessions, freeing hardware resources and reducing the risk of interrupted work. The ability to run many more concurrent agents expands experimentation capacity, which can accelerate development cycles. Cost control will depend on careful monitoring of instance usage, especially when scaling beyond the Micro tier. Finally, the pending governance feature signals that enterprises should plan for interim policy enforcement mechanisms while awaiting Docker’s native solution.