Show HN: Skill that lets Claude Code/Codex spin up VMs and GPUs
CloudRouter introduces a new primitive for AI agents, allowing them to spin up cloud VMs and GPUs for development and browser automation directly. This tool tackles the chaotic resource sharing of local machines by providing isolated, pre-configured environments with integrated dev tools. Its popularity stems from simplifying complex cloud provisioning and enabling more robust, parallel AI-driven development workflows.
The Lowdown
CloudRouter is presented as a skill and CLI tool designed to empower AI coding agents, such as Claude Code and Codex, with the capability to provision and manage cloud-based virtual machines and GPUs. Traditionally, AI agents performing coding tasks would operate on a local machine, leading to resource contention and a lack of necessary tools like browsers or dedicated GPUs. CloudRouter aims to solve this by abstracting the complexities of cloud infrastructure.
- It enables agents to initiate isolated VMs directly from local project directories, complete with file synchronization and the ability to execute commands.
- VMs come pre-configured with essential development tools like VNC desktops, VS Code, and Jupyter Lab, accessible via authenticated URLs.
- A significant feature is the on-demand provisioning of GPUs, ranging from T4s to B200s, allowing agents to execute ML training and inference tasks without manual setup.
- The tool supports comprehensive browser automation, enabling agents to navigate, interact with elements, and take screenshots within the VM's browser.
- CloudRouter inverts the typical cloud development workflow, keeping the agent local while pushing its operational environment to the cloud, thus facilitating parallel task execution.
- It is open-source, built in Go, and distributed as an npm package, providing flexibility and transparency.
By providing a streamlined interface for cloud resource management, CloudRouter offers a powerful solution for enhancing the autonomy and efficiency of AI coding agents, particularly for tasks requiring dedicated environments, browser interaction, or significant computational power.
The Gossip
Seamless Sandboxes vs. Cloud CLIs
The discussion immediately turned to CloudRouter's unique value proposition compared to simply having AI agents interact with raw AWS, Azure, or GCP CLI tools. The author clarified that CloudRouter abstracts away significant complexity, offering a "nice primitive" that provides pre-baked environments with SSH, file synchronization, browsers, GPUs, Docker, VNC, and Jupyter Lab, thereby bypassing the need to manage cloud account setup, security groups, and intricate environment configurations.
Bug Bash & Swift Solutions
Early adopters quickly began testing CloudRouter, with one user reporting an unexpected password request during initial setup. The author demonstrated impressive responsiveness, identifying the root cause and pushing a fix shortly after the issue was reported, highlighting active development and a commitment to a smooth user experience.
Awe and Admiration
Initial comments expressed strong positive sentiment, praising the demo and the concept itself. Users found the project "awesome" and an "enjoyable read," indicating a strong positive first impression and interest in the capabilities offered by CloudRouter.