Welcome to Herd: The Application Delivery Plane
Herd is a high-performance Application Delivery Plane designed to bridge the gap between low-level microVM hardware and the modern application developer experience.
Built on Firecracker, Herd automates the process of translating standard Docker/OCI images into isolated, session-affine compute environments, complete with built-in L7 ingress and zero-config networking.
⚡ The Four Delivery Layers
Herd's architecture is built on four critical operational layers that guarantee speed, security, and ease of use:
- OCI Translation: Pulls images via
containerd, extracts execution metadata (CMD,ENV), and automatically provisions Copy-on-Write rootfs snapshots. - L7 "Wake-on-Request" Proxy: A high-speed reverse proxy that intercepts HTTP requests, cold-boots the corresponding microVM in sub-500ms (if needed), and tunnels traffic securely into the guest.
- Automated IPAM: A zero-config networking engine that manages PTP TAP interfaces, deterministic internal IP allocation, and host-side NAT routing.
- Guest Agent Execution: Injects a meticulously crafted, static
herd-guest-agentas PID 1, which bootstraps internal microVM networking and manages the user workload execution.
🚀 Navigation
Get started with Herd today:
- Getting Started: Installation and your first deployment.
- Architecture Deep Dive: Understanding the magic beneath the hood.
- CLI Reference: Mastering the
herdcommand suite. - Configuration: Deep dive into
herd.yaml.