Argo CD is a declarative, GitOps continuous delivery tool for Kubernetes. It keeps a cluster’s live state in sync with the desired state defined in a Git repository, so deployments become a matter of merging a pull request rather than running imperative commands against a cluster.

How it works

Argo CD runs as a controller inside a Kubernetes cluster. It continuously compares what’s declared in one or more Git repositories against what’s actually running, and can either surface drift or reconcile it automatically. Because the desired state lives in Git, every change is versioned, reviewable, and revertible through the same pull request workflow teams already use for code.

Core concepts

• Application — the basic unit Argo CD manages: a path in a Git repository mapped to a destination cluster and namespace.
• Sync policy — controls whether Argo CD applies changes automatically as soon as Git changes, or waits for a manual sync.
• Health and sync status — Argo CD tracks both whether resources are healthy (e.g. pods running) and whether they match Git (in sync, out of sync, or unknown).
• Projects — group applications together and restrict which repositories, clusters, and resource kinds they’re allowed to use.

Why teams adopt GitOps with Argo CD

• Auditable by default. Every change to the cluster has a corresponding commit and, typically, a reviewed pull request.
• Easy rollback. Reverting a deployment is a git revert, not a manual kubectl intervention under pressure.
• Consistent across environments. The same reconciliation model applies whether you’re targeting a single namespace or dozens of clusters.
• Drift detection. Changes made directly against a cluster show up immediately as “out of sync,” instead of silently diverging from what’s documented in Git.

Getting started

Install Argo CD into a Kubernetes cluster, point it at a Git repository containing your manifests — plain YAML, a Helm chart, or a Kustomize overlay all work — and define an Application resource that ties the two together. The guides below walk through the most common setups, from a single application to managing multiple clusters.