All articles

GitOps Tools in 2026: Argo CD, Flux, Fleet, OpenShift GitOps, Kargo and How to Choose

Argo CD, Flux, Fleet, OpenShift GitOps, GitLab and Kargo compared: what each tool does, where it fits, and how to choose a GitOps tool in 2026.

Search for "GitOps tools" and most answers turn into a two-way race between Argo CD and Flux. That is a fair place to start: both are CNCF graduated projects, and both run the reconciliation loop that GitOps depends on. But the field in 2026 is wider than two engines. There are distributions built on them, tools for running GitOps across many clusters, and a separate layer for promoting changes between environments. This article maps the six names that come up in almost every evaluation, what each one is for, and how to choose. If you are new to the model itself, start with our guide to GitOps.

Key takeaways

  • A GitOps tool is an agent that pulls the desired state from Git and keeps reconciling the cluster to it; a CI server on its own is not one.
  • Argo CD and Flux are the two engines, and most other GitOps tools package, scale or extend one of them.
  • Argo CD's 2025 user survey shows deep use, and Flux remained a CNCF graduated project after its main sponsor shut down.
  • Fleet targets many clusters, OpenShift GitOps packages Argo CD for OpenShift, GitLab pairs Flux with its agent, and Kargo adds promotion between environments.
  • Choose by platform, cluster count, who needs a UI and how versions move between stages, then pick what your team can operate.
The DevOps loop of plan, build, continuous integration, deploy and operate, joined by continuous feedback. GitOps tools run the deploy and operate half of the loop.

What counts as a GitOps tool?

The OpenGitOps principles define four properties of a GitOps-managed system: its desired state is declarative, versioned and immutable, pulled automatically and continuously reconciled. The last two are what a tool adds. In the principles' own words, "Software agents automatically pull the desired state declarations from the source", and they "continuously observe actual system state and attempt to apply the desired state." A GitOps tool is that agent. It runs in or next to your cluster, watches a Git repository and keeps correcting the difference.

Jenkins and Terraform are not GitOps tools on their own

A CI job that runs kubectl apply or terraform apply after a merge is push-based deployment: nothing watches the target afterwards, so a manual change goes unnoticed. Both can feed a GitOps setup, by building images or provisioning clusters, but the pulling, reconciling agent is what makes it GitOps.

Which GitOps tools are there in 2026?

Two of the six are engines. The other four package an engine, run one across many clusters or add a step around it.

ToolWhat it isA good fit when
Argo CDA GitOps engine for Kubernetes with a web UI. Its ApplicationSet controller lets one manifest target many clusters (Argo CD docs).People need to see and act on application state, and many teams share one platform.
FluxA set of Kubernetes controllers built on the GitOps Toolkit, described by the CNCF as an "Open and extensible continuous delivery solution for Kubernetes".You want a lean, composable engine that other automation drives.
FleetCustom resources and controllers from the Rancher project that deploy raw YAML, Helm charts or Kustomize from Git (Fleet docs).You run many clusters, especially under Rancher.
Red Hat OpenShift GitOpsAn Operator that uses Argo CD as its GitOps engine, with extra automation and OpenShift integration (Red Hat docs).You run OpenShift and want Argo CD packaged for the platform.
GitLab with FluxGitLab integrates Flux, and its agent for Kubernetes adds cluster access management and the cluster state in the GitLab UI (GitLab docs).Your code and pipelines already live in GitLab.
KargoA promotion layer that complements Argo CD, built by the creators of Argo (Kargo).Versions have to move through environments in a fixed order.

How widely used is Argo CD, and is Flux still a safe choice?

The CNCF and the Argo maintainers surveyed Argo CD's production users in 2025 (CNCF, July 2025):

~60%

of respondents' Kubernetes clusters run Argo CD for application delivery

97%

of respondents use it in production, up from 93% in 2023

42%

run more than 500 applications per Argo CD instance, up from 15% in 2023

These figures come from Argo CD's own users, so they show depth of use rather than market share. Flux has no comparable survey, but its position is sound. It graduated in the CNCF on November 30, 2022, days before Argo did on December 6 (CNCF project pages for Flux and Argo). It also outlived its main sponsor: when Weaveworks shut down its commercial operations in early 2024, GitLab publicly recommitted to the project, calling Flux "a mature Cloud Native Computing Foundation (CNCF) project with a large and dedicated user base" (GitLab, March 2024).

For a feature-by-feature comparison of the two engines, see our article GitOps in 2026: reconciliation, drift, and the Argo CD vs Flux decision. The short version still holds: Argo CD when people need to see state, Flux when the audience is other automation.

What do Fleet, OpenShift GitOps and GitLab add?

None of the three replaces the reconciliation loop. Each solves a problem around it: scale, a platform-specific distribution, or one place for code and clusters.

Fleet: many clusters

Kubernetes custom resources and controllers that manage GitOps for one cluster or a large fleet of them. Every source, whether raw YAML, Helm or Kustomize, is turned into a Helm chart before it is deployed.

OpenShift GitOps: Argo CD, packaged

An Operator that uses Argo CD as the GitOps engine and enables GitOps workflows across multicluster OpenShift and Kubernetes infrastructure.

GitLab: Flux plus an agent

GitLab integrates Flux for GitOps and recommends running it with its agent for Kubernetes, which simplifies the Flux setup, manages cluster access and shows the cluster state in the GitLab UI.

What is Kargo, and do you need a promotion tool?

Argo CD and Flux keep each environment in line with what Git says. They do not decide when a version that passed in test should move to staging and then to production. Most teams script that step in CI, with a pipeline that edits the next environment's configuration once checks pass. Kargo makes the step explicit. It describes itself as "a continuous promotion orchestration layer, that complements Argo CD for Kubernetes", which automates the promotion of changes from one environment to the next and aims to remove the need for custom automation or CI pipelines to do it (Kargo). Whichever tool owns it, a promotion flow has the same parts:

  1. A new version appears

    A new image tag or chart version becomes a candidate for release.

  2. It lands in the first environment

    The version is written to that environment's configuration in Git, and the GitOps engine syncs it.

  3. Checks decide whether it moves on

    Tests, metrics or a person verify the environment before the same version becomes eligible for the next stage.

  4. Production gets exactly what was verified

    The version that passed is promoted unchanged, and Git records who promoted what, and when.

If you deploy a handful of services to two environments, a pipeline step is enough. With dozens of services and several stages, a promotion layer saves you from maintaining the same logic in every pipeline.

How do you choose a GitOps tool?

Start from your constraints rather than feature lists. Both engines are free, open source and CNCF graduated, so the real cost is running, securing and upgrading them.

  • Your platform comes first. On OpenShift, OpenShift GitOps is Argo CD packaged for the platform. If you already run Rancher, look at Fleet, which comes from the same project.
  • Count clusters, not applications. For one cluster or a few, either engine works. For dozens, plan for Argo CD ApplicationSets or Fleet from the start.
  • Decide who needs a UI. Argo CD ships a web UI. For Flux, user interfaces are separate projects, and GitLab shows cluster state in its own UI.
  • Make promotion an explicit choice. Decide whether a pipeline, a script or Kargo moves versions between environments, and who may approve it.
  • Pick what your team can operate. A GitOps engine is a production system with its own upgrades, access control and secrets. The best tool is the one someone owns.

Where to start

Shortlist one engine, run it against a non-production cluster without automatic sync, and list every difference between Git and the cluster before you turn sync on. Add ApplicationSets or Fleet when the number of clusters grows, and a promotion layer when the number of environments does. Our GitOps guide walks through a first rollout in five steps.

Sources

  1. CNCF End User Survey Finds Argo CD as Majority Adopted GitOps Solution for Kubernetes - Cloud Native Computing Foundation, July 24, 2025
  2. OpenGitOps: GitOps Principles - OpenGitOps
  3. Introduction to ApplicationSet controller - Argo CD documentation
  4. Flux - Cloud Native Computing Foundation, project page
  5. Argo - Cloud Native Computing Foundation, project page
  6. The continued support of FluxCD at GitLab - GitLab, March 5, 2024
  7. Using GitOps with a Kubernetes cluster - GitLab Docs
  8. Overview - Fleet documentation
  9. About Red Hat OpenShift GitOps - Red Hat Documentation, Red Hat OpenShift GitOps 1.16
  10. Kargo - Akuity

Where to start

Start with your last release

The free Release Check shows where your last release spent its time and what that costs a year, in under 5 minutes and without a call.

Keep reading