> ## Documentation Index
> Fetch the complete documentation index at: https://radarhq.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# OSS vs Cloud

> How Radar OSS and Radar Cloud differ - features, deployment model, and when to pick which.

Radar comes in two flavors. The cluster-side binary is **identical** in both. Cloud adds a hosted, multi-tenant control plane on top: orgs, RBAC, SSO, billing, audit log, fleet views. **No features are hidden behind Cloud** - if it's documented anywhere in these docs, it ships in the OSS binary.

## Side by side

|                            | **Radar OSS**                                         | **Radar Cloud**                                                                        |
| -------------------------- | ----------------------------------------------------- | -------------------------------------------------------------------------------------- |
| **License**                | Apache 2.0                                            | Proprietary SaaS (engine remains Apache 2.0)                                           |
| **Where it runs**          | Your machine or your cluster                          | Hosted control plane at `app.radarhq.io`; the Radar binary still runs in your cluster  |
| **Account required**       | No                                                    | Yes (WorkOS auth)                                                                      |
| **Cost**                   | Free                                                  | Free tier (3 clusters) / Team / Enterprise - see [pricing](https://radarhq.io/pricing) |
| **Multi-cluster**          | One Radar at a time; switch by re-pointing kubeconfig | All clusters under one URL, fleet views                                                |
| **Auth**                   | OIDC or proxy headers, configured per install         | WorkOS (Google / passkey / SSO), centrally                                             |
| **Team workflow**          | Whatever auth you wire in                             | Orgs, roles, invitations, audit log                                                    |
| **Persistence**            | Pod-local (timeline in memory or SQLite)              | Postgres for org / billing / audit; pod-local for cluster data                         |
| **MCP for AI**             | Local MCP endpoint                                    | Per-cluster MCP via personal access tokens                                             |
| **Cross-cluster topology** | Per-cluster only                                      | Per-cluster today; cross-cluster on the roadmap                                        |
| **SSO (SAML / OIDC)**      | OIDC for the web UI                                   | SAML, Google, Okta via WorkOS Admin Portal                                             |
| **Billing**                | Not applicable                                        | Stripe, per-cluster on Team plan                                                       |
| **Audit log**              | Pod-local Kubernetes events only                      | Org-scoped activity log with CSV export                                                |
| **Notifications**          | None (emit your own from K8s events)                  | In-app inbox + Slack / webhooks for cluster / org / billing events                     |

## Which one fits you

<CardGroup cols={2}>
  <Card title="Pick OSS if you..." icon="server" href="/docs/install">
    * Want a fast, ad-hoc replacement for `kubectl` and the K8s dashboard, on your own machine.
    * Need to run Radar in an air-gapped environment.
    * Are comfortable wiring auth (oauth2-proxy, OIDC) yourself for a shared in-cluster install.
    * Prefer self-hosted with no vendor in the loop.
  </Card>

  <Card title="Pick Cloud if you..." icon="cloud" href="/docs/quickstart-cloud">
    * Manage more than one cluster and want them all behind a single URL.
    * Have a team and want central role-based access, invitations, and an audit trail.
    * Want SSO without configuring OIDC per cluster.
    * Want to give Claude / Cursor / Claude Code MCP access to your clusters without managing tokens per cluster.
    * Want fleet-wide views: which clusters have audit failures, which packages are drifting, where a resource lives across clusters.
  </Card>
</CardGroup>

## Use both at once

Run OSS locally against your dev cluster while your team uses Cloud for shared environments. Same binary; the only difference is whether `cloud.enabled` is set in Helm.

## Switch later

* **OSS → Cloud**: install the chart with `cloud.enabled=true` against the same cluster.
* **Cloud → OSS**: drop the `cloud.*` Helm values and Radar reverts to a self-contained in-cluster install.

K8s is the source of truth in both cases - there's no Radar-side data format to migrate.