Descriptions:
Web Dev Cody walks through how he dramatically improved sandbox provisioning speed for his Mission Control platform by replacing runtime bootstrap scripts with pre-built AMI images using HashiCorp Packer.
The original setup spun up AWS EC2 instances and ran a user-data script each time to install Claude Code, OpenCode, and other required tools — a slow process that could also yield different dependency versions across sandboxes, making environments non-reproducible. By using Packer, Cody now automates the creation of a snapshot image that pre-installs all tooling on a T3 Medium instance, copies the resulting AMI across multiple AWS regions, and publishes a versioned manifest.json so that Mission Control always pulls the latest verified image. A GitHub Actions workflow triggers the full build-and-publish pipeline on each tagged release in the public MC Sandbox repository.
The video walks through the actual orchestrator script (buildAMI.mjs), the Packer HCL configuration file, the install.sh provisioning script, and the manifest publishing step in detail. Cody also addresses the security transparency angle — making the image and install scripts fully public so users can audit exactly what gets baked into the sandbox environment. The end result cuts provisioning time down to roughly 30 seconds to one minute per sandbox while guaranteeing a consistent, reproducible environment on every deployment — a meaningful improvement for any developer building multi-user AI coding sandbox infrastructure on AWS.
📺 Source: Web Dev Cody · Published June 08, 2026
🏷️ Format: Hands On Build
