ArgoCD Permissions
You set up argocd permissions by mapping identity provider groups to role rules. These rules control who can read, sync, or delete apps inside your clusters. Most teams start by linking their single sign-on groups to built-in admin roles. Overly broad roles let any developer change production state or view cluster secrets. How argocd permissions work in practice The platform uses role-based rules to check every user request at the server level. Admins write policy lines that link a subject to
Your coding agents run with your access.
Claude Code reaches prod with a developer's credentials, and approval prompts get clicked through. hoop.dev checks every query, masks sensitive data, and holds risky changes for a human.
You set up argocd permissions by mapping identity provider groups to role rules. These rules control who can read, sync, or delete apps inside your clusters. Most teams start by linking their single sign-on groups to built-in admin roles. Overly broad roles let any developer change production state or view cluster secrets.
How argocd permissions work in practice
The platform uses role-based rules to check every user request at the server level. Admins write policy lines that link a subject to a specific resource and action. The subject is usually a team group from your identity provider or a local user. The resource is an app, a cluster, a repo, or a project group. The action is a verb like get, create, update, sync, override, or delete. When a user logs in, the system reads their group claims from the token. Then it checks the policy map to see if that group can run the action.
Projects add another boundary around your apps and clusters. A project groups related apps together and sets rules for where they can deploy. You can lock a project so it only deploys to one specific cluster namespace. You can also limit which source repos a project can pull manifests from. Each project can hold its own local roles and issue tokens for automation pipelines. This helps platform teams delegate app sync rights without giving away full cluster admin keys.
How to scope roles for teams and projects
A clean setup starts with a strict default role for all logged-in users. Most teams set the default role to read-only across non-sensitive projects. You then create specific roles for each product team in your company. A team role grants sync and update rights only to apps inside their assigned project. Admins keep the rights to add clusters, register repos, and edit global system settings.
You should also avoid local user accounts for daily human tasks. Map your single sign-on groups directly to project roles instead. When an engineer leaves a team, updating the identity group removes their deploy rights right away. Keep local accounts and project tokens strictly for continuous integration jobs that trigger syncs.
ArgoCD Security + AI Agent Permissions: architecture patterns and best practices
Continue reading? Get the full guide.
Free. No spam. Unsubscribe anytime.
What native role rules solve and where they fall short
Native role rules do a great job of sorting out who can see which app. They stop a developer on one team from syncing another team's service by mistake. They also block unauthorized users from adding rogue git repos to the control plane. For basic day-to-day dashboard use, these project boundaries keep teams in their own lanes.
Still, static role maps only check who logs in and which button they click. They do not inspect the actual changes inside a sync or an override request. If an engineer has sync or update rights for a production emergencies role, that right stays active all day. A broad role cannot require a peer review before a destructive sync runs in production. Also, when developers use the pod exec feature or view live logs, native rules cannot mask sensitive fields. Emails, phone numbers, card numbers, and national IDs in pod output appear in plain text. Closing this gap requires control over individual requests and actions in flight.
Add request guardrails and response masking beside Kubernetes
Cluster roles and service accounts govern who reaches the Kubernetes API. Still, a broad role or agent can delete workloads or print sensitive values from ConfigMaps and logs. In practice, hoop.dev runs beside the API server as a sidecar for your Kubernetes API and checks every request and response alongside RBAC.
With hoop.dev Kubernetes controls, guardrails can allow listing Secret names while refusing reads of Secret data before execution. Data masking redacts matched keys in returned JSON responses and log streams in memory. Furthermore, hoop.dev holds SOC 2 Type II and is MIT licensed on GitHub.
Frequently asked questions
How do I stop users from seeing all apps by default?
Set the global default policy role to an empty value or a minimal role. Then grant read and sync rights only through specific group mappings tied to individual projects.
Should I use local users or single sign-on groups for team access?
Always use single sign-on groups for human engineers so access updates match team changes. Save local accounts and project tokens for automated build pipelines that need to trigger syncs.
Can project roles stop a developer from reading secret values in pod logs?
No, if a role allows log viewing, the user sees the entire log stream. You need a separate data masking rule in flight to rewrite emails or card numbers in results.
Related guides
In this post