I built a self-hosted Fly.io engine using Go and Firecracker
Firecracker microVMs from Docker images with built-in L7 proxy and wake-on-request.
microVM orchestrators to boot from docker images, packed with L7 reverse proxy
Firecracker microVMs with Docker CLI UX, though Kata Containers already solves container isolation.
DevOps engineers, security-conscious developers, backend infrastructure teams
Kata Containers · gVisor · KubeVirt
Lately, I have been trying to understand the security aspect of docker containers, and what I have realized is that all docker containers share the host's kernel. Any zero day vulneerability in the kernel can be used to gain access to the host os.
In order to deal with this, I did some research turns out Amazon has open sourced the core technology behind their serverless technology lambda. But in its current state its very hard to setup, let alone run anything securly. This technology is called firecracker microvm's
It started off as a go library, for creating process pools to just do a simple firecracker spawn, turned into a full fledged host side daemon.
deploying a microvm through an image is now as simple as running
`herd deploy --image postgres:latest -p 5432:5432 -e POSTGRES_PASSWORD=postgres`
with boot times ~500ms
That brings us to today. I am looking for people to test this out and provide some feedback, I have been warned/cautioned by a lot of friends that building in isolation is a recipe for disaster.
PS: it only works on linux, macos doesn't have the required isolation, and I stopped caring about winslop.
Firecracker microVMs from Docker images with built-in L7 proxy and wake-on-request.
Firecracker microVM sandbox for agents in 5 seconds, Claude Desktop ready.
Hardware-isolated VM sandbox for Claude, 2-second boot, no Docker complexity.
Firecracker microVMs for agent isolation but Linux-only with KVM requirements.
Auto-generates API keys to block the 175k exposed Ollama instances.
Auto-generated tokens block the 175k exposed Ollama servers found online.