FlukeGitOps-based continuous delivery for VMs and bare metal

A single-binary reconciliation-loop CD tool for long-lived servers. No Kubernetes required.

New to Fluke?

Follow the tutorial to get a server, agent, and your first manifest running.

Getting Started

Step-by-step guides for common tasks.

How-to Guides

Complete reference for manifests, config, executors, and CLI.

Reference

Architecture, executor model, and the reasoning behind key decisions.

Explanation

Git repo (desired state)
       ↓  poll / webhook
    fluke server
       ↓  gRPC
    fluke agent   ←→   mise, systemd, shell...
   (on each host)

Single binary — server and agent modes in one binary, selected by flag
HCL manifests — expressive desired state with per-executor schemas
Pluggable executors — mise, systemd, shell built-in; designed for external plugins later
Label-based targeting — match hosts by labels, not hostnames
Configurable drift handling — remediate silently, alert-and-remediate, or alert-only
Intentionally stateless — in-memory event history by default, Redis opt-in