Skip to main content
Ecosystem & Marketplaces›Open-Source MCP Servers (10)›ArgoCD MCP Server

ArgoCD MCP server

The ArgoCD MCP Server is a Model Context Protocol server for GitOps continuous delivery on Kubernetes. It exposes 29 tools, real-time MCP resources (argocd://), and operational prompts covering the complete ArgoCD v2.x lifecycle—repository onboarding, AppProject multi-tenancy, application creation, manifest diffing, dry-run synchronization, health diagnostics, and revision rollbacks.

29 GitOps Tools

Manage applications, preview live vs. desired manifest diffs, trigger dry-run or live syncs, onboard repositories, and configure RBAC projects.

Write Protection & Config

Default to read-only diagnostics (MCP_ALLOW_WRITE=false) and keep ArgoCD API tokens and Git credentials isolated in server environment variables.

Resources & Guided Prompts

Stream live application health (argocd://applications), sync metrics, and multi-step deployment/rollback playbooks.

Production Examples

End-to-end patterns for onboarding GitHub repositories, debugging OutOfSync or Degraded apps, and executing emergency rollbacks.

Core capabilities​

DomainCapabilities
Application LifecycleCreate, update, validate, and delete ArgoCD Application resources; inspect per-resource health, sync status, and Kubernetes event streams.
Sync & Rollback OperationsPreview live vs. target Git manifest diffs, run dry-run syncs, prune orphaned resources, hard/soft refresh caches, cancel in-flight operations, and roll back to previous revisions.
Repository OnboardingRegister and validate HTTPS or SSH Git repositories using server-side credentials that are never exposed to the LLM context window.
Projects & Multi-TenancyCreate AppProject resources with scoped source repositories, destination clusters/namespaces, resource allow/deny lists, and RBAC roles.
Diagnostics & Log AnalysisRetrieve pod logs with automatic error-line filtering, correlate failed sync hooks, and inspect cluster-wide GitOps health metrics.

Architecture​

Quickstart​

1
Prepare your ArgoCD API token

Ensure you have a running ArgoCD v2.x instance and an authentication token (ARGOCD_AUTH_TOKEN).

2
Start the ArgoCD MCP Server container

Run the Docker image and pass your ArgoCD server URL and credentials via environment variables:

docker run --rm -it \
  -p 8770:8770 \
  -e ARGOCD_SERVER_URL="https://host.docker.internal:8080" \
  -e ARGOCD_AUTH_TOKEN="your-argocd-token" \
  -e MCP_PORT=8770 \
  -e MCP_PATH="/mcp" \
  -e ARGOCD_INSECURE="true" \
  -e MCP_ALLOW_WRITE="false" \
  talkopsai/argocd-mcp-server:latest

Use host.docker.internal:8080 when connecting to a locally port-forwarded ArgoCD server (kubectl port-forward svc/argocd-server -n argocd 8080:443). Set MCP_ALLOW_WRITE="true" only when you want to permit sync, create, or rollback mutations.

3
Connect your MCP client

Register the server URL in your MCP client configuration:

mcp.json
{
  "mcpServers": {
    "argocd-mcp-server": {
      "url": "http://localhost:8770/mcp",
      "disabled": false
    }
  }
}

Security model​

  • Read-only by default: Keep MCP_ALLOW_WRITE=false for observability and incident triage so agents can inspect diffs, events, and logs without mutating GitOps state.
  • Server-side credential isolation: Git tokens (GIT_PASSWORD, SSH_PRIVATE_KEY_PATH) and ARGOCD_AUTH_TOKEN remain inside the MCP server process and are never required as LLM tool arguments.
  • Dry-run verification: Always preview manifest diffs (get_application_diff) and run sync_application(dry_run=True) before applying production changes.

Project structure​

argocd-mcp-server/
├── argocd_mcp_server/
│   ├── tools/           # 29 MCP tools (apps, deploy, repos, projects, diagnostics)
│   ├── resources/       # Real-time argocd:// resource streams
│   ├── prompts/         # Guided deployment, onboarding, and rollback workflows
│   ├── services/        # Async ArgoCD REST/gRPC client services
│   ├── server/          # FastMCP server setup
│   └── utils/           # Validation and formatting helpers
├── scripts/             # Token helper utilities (fetch_argocd_token.py)
├── tests/               # Unit and integration tests
└── pyproject.toml       # Project configuration
On this page