🔐 Authentik + MCP Integration: Full API Access & Safe Diagnostics

[cdmx-in]

Overview

We’ve successfully integrated Authentik with the Model Context Protocol (MCP) using two server modes:

• authentik-mcp: Full-featured API access
• authentik-diag-mcp: Read-only diagnostics mode

This setup powers secure automation, streamlined monitoring, and developer productivity across Authentik environments.

---

✨ Key Features

• Dual Operation Modes:

  • authentik-mcp: Full CRUD capabilities for users, groups, applications, flows, and more.
  • authentik-diag-mcp: Read-only access ideal for monitoring, logging, and audits.

• Developer-Friendly Tooling:

  • Standardized MCP calls to list users, fetch logs, inspect events, and more.
  • Seamlessly integrates with tools like VS Code, Claude Desktop, and CI/CD workflows.

• Effortless Setup:

  • Minimal configuration with a single entry point.
  • Secure token-based access for quick onboarding.

---

🧪 Example: VS Code + GitHub Copilot MCP Extension

settings.json

"mcp": {
  "servers": {
    "authentik": {
      "command": "uvx",
      "args": [
        "authentik-diag-mcp",
        "--base-url", "https://your-authentik-instance",
        "--token", "your-api-token"
      ]
    }
  }
}

claude_desktop_config.json

{
  "mcpServers": {
    "authentik": {
      "command": "uvx",
      "args": [
        "authentik-diag-mcp",
        "--base-url",
        "https://your-authentik-instance",
        "--token",
        "your-api-token"
      ]
    }
  }
}

Use the MCP UI to:

• List users
• Fetch audit logs
• Monitor event flows

Sample MCP Commands:

• authentik_list_users_info (diagnostic)
• authentik_list_users (full API)

---

💡 Why It Matters

• Safe by Design: Read-only mode prevents accidental changes during diagnostics.
• Automation-Ready: Ideal for CI/CD pipelines, observability tools, and workflow automation.
• Open & Extendable: Built for teams, contributors, and communities.

---

🙏 Credits

Thanks to the incredible contributors in the Authentik and MCP ecosystems!

🔗 View on GitHub