Firecracker

Firecracker is a virtual machine monitor (VMM) that uses Linux's KVM to create and run microVMs. It was built at AWS to power services like Lambda and Fargate, and it's aimed at anyone building multi-tenant container or function platforms that need hardware-level isolation without the overhead of a full VM. If you operate infrastructure where you run untrusted or third-party code and need fast startup times with a small attack surface, this is the layer that sits between your orchestrator and the kernel.

Key features

• Minimalist device model: excludes unnecessary devices and guest-facing functionality, which cuts memory footprint and attack surface compared to general-purpose VMMs.
• API-driven configuration: a single VMM process exposes an OpenAPI-specified endpoint for setting vCPU count, memory size, CPU templates, network interfaces, and disks.
• Rate limiting: virtio devices can be configured with bandwidth and IOPS limits per device.
• Live disk management: block device backing files can be swapped and rescanned before or after boot, so the guest picks up size changes.
• vsock, entropy, and pmem device support: additional virtio devices for host-guest communication, randomness, and persistent memory.
• Memory hotplugging: memory can be added or resized on a running microVM.
• Device hot-plug (developer preview): virtio PCI devices can be added or removed while the VM runs.
• Jailer process: applies cgroup/namespace isolation and drops privileges before starting Firecracker, intended for production deployments.
• Demand fault paging and CPU oversubscription: enabled by default to improve density on the host.
• Seccomp filtering: thread-specific filters restrict syscalls available to each Firecracker process.

Ideal use cases

Firecracker fits when you're building a serverless platform, container-as-a-service offering, or any multi-tenant system that needs to run untrusted code with strong isolation and fast (sub-second) startup. It's already integrated into runtimes like Kata Containers and Flintlock, so it also works as a lower-level building block for existing container tooling that wants VM-grade isolation instead of namespace-only isolation.

It's not the right choice if you need a desktop virtualization tool, GPU passthrough, full device emulation (USB, audio, graphics), or a VM that behaves like a general-purpose server with a broad set of peripherals. Firecracker deliberately strips out anything not needed for container/function workloads, so workloads expecting a full-featured guest environment (or Windows guests) won't be served well by it. It also depends on KVM, so it only runs on Linux hosts with virtualization support, and production-grade security depends on a properly hardened host OS.

Installation

You can grab a prebuilt binary from the releases page, or build from source. Building requires a Unix/Linux system with Docker and bash installed; the project uses a development container to standardize the build environment.

git clone https://github.com/firecracker-microvm/firecracker
cd firecracker
tools/devtool build
toolchain="$(uname -m)-unknown-linux-musl"

After building, the binary lands at build/cargo_target/${toolchain}/debug/firecracker. From there, follow the quickstart guide for running your first microVM (you'll need a kernel image and root filesystem). For anything beyond local testing, review the production host setup document, since Firecracker's security guarantees depend on how the host OS is configured. If you plan to run it in a multi-tenant production environment, also look at the Jailer, which wraps the Firecracker process with cgroup/namespace isolation and privilege dropping.