Overview
The Druva Model Context Protocol (MCP) server enables compatible AI agents to securely connect to the Druva Cloud using your existing account permissions.
By leveraging this gateway, you can interact with your cloud environment using natural language on supported AI tools and interfaces.
Important
Setup requirements and approval workflows vary by AI agent. The Druva MCP server currently has read and modify permissions, but it cannot delete Druva Cloud configuration or data.
Key Benefits
Enhanced Operational Efficiency: Administrators can monitor backup fleets, inspect historical audit logs, and pull legal discovery states without context-switching to the primary Druva administration console.
Governed Security and RBAC Layer: All interactions are strictly constrained by implicit Role-Based Access Control (RBAC) via a 3-Legged OAuth 2.1 flow. The AI application acts exclusively on behalf of the authenticated user; if an operator lacks console permissions to view a specific workload, the native API blocks execution.
How Druva MCP Server Works
The Druva MCP server establishes a secure remote HTTP connection:
You add the Druva MCP server URL to the supported AI agent.
The agent prompts you to authenticate via the Druva Console.
Once authenticated, the Druva MCP server exposes supported tools to your agent.
The agent uses these tools to fetch environment data and answer your natural language requests.
Compatibility- Supported AI Agents
The Druva MCP server currently supports following AI Agents:
Claude ( Desktop, Code, Coworks, Web)
Cursor/ VS Code
Configure Druva MCP server
Step 1: Configure the remote Druva MCP server
To connect an MCP-compatible agent to Druva, add the Druva MCP server using the following server details.
Name: Druva MCP (or any name that helps you identify the server.)
URL:
For APAC region customers: https://mcp-ap.druva.com/mcp
For US region customers: https://mcp-us.druva.com/mcp
📝 Note
The Druva MCP server is currently not available on Druva Australia control plane and GovCloud environments.
Example agent configuration
Use the format supported by your MCP-compatible agent. The exact fields and setup steps vary by agent.
JSON example:
{
"mcpServers": {
"druva": {
"type": "http",
"url": "https://mcp-ap.druva.com/mcp"
}
}
}
TOML example:
[mcp_servers.druva]
url = "https://mcp-ap.druva.com/mcp"
Confirm the required configuration format in the documentation for your MCP-compatible agent.
Step 2: Authentication
When you connect to the MCP server for the first time, your agent prompts you to authenticate by redirecting to the Druva console. Use the same credentials that you use to sign in to the Druva Console. Authentication supports both password and TOTP methods.
Step 3: Test the connection
Start your MCP-compatible agent after you add the Druva MCP server.
Ask a simple Druva question, such as:
List available Druva tools.
The connection is working if the agent uses the Druva MCP server and returns a Druva response that matches your permissions.
Example Operational Prompts
Once connected, you can monitor your Druva environment using natural language commands:
"Provide a summary of the global backup success rate across all regions."
"Create a daily health overview detailing total data backed up, success percentages, and active critical alerts."
"Show me the resources which are consuming the most space and what kind of files are stored on them."
"Check the backup health status for all integrated SaaS applications."
"Look at the audit logs and identify any unusual activities by admin [Name]."
"Which users are currently on legal hold and how much data is backed up for those users?"
“Produce a detailed backup health report for all endpoints devices”
Druva MCP Server Capabilities
Druva MCP server capabilities are available through tools and skills. This MCP server includes the following tools.
Tool | Purpose |
| Lists the Druva skills or capabilities available through the Druva MCP server. The AI agent uses this tool to discover which Druva actions are supported. |
| Identifies the most relevant Druva skill for a user request. The agent uses this tool when a natural-language request must be matched to a supported Druva capability. |
| Retrieves the details needed to use a selected Druva skill. The agent uses this tool after it selects a skill and needs the instructions or action details required to continue. |
| Runs the supported script or action associated with the selected Druva skill. The agent uses this tool to fetch data from the Druva environment. |
Limitations & Access Control
Unsupported APIs: The Event API is not yet compatible with the MCP server and will be available soon.
SSO Restrictions: MCP server is currently not supported for Administrators using SSO authentication and will be available soon. Authentication supports both password and TOTP methods.
Agent-Side Approvals: Some agents ask you to manually approve tool execution before a script runs. This is an agent-specific safety feature, not a Druva-controlled setting.
Interpretation Variance: Results can vary slightly depending on how your specific AI agent interprets and phrases background prompts.
Troubleshooting & FAQs
Why can’t my MCP-compatible agent connect to the Druva MCP server?
Verify the following:
The server URL is exactly the one mentioned below -
APAC Region: https://mcp-ap.druva.com/mcp
US Region: https://mcp-us.druva.com/mcp
The agent supports remote HTTP-based MCP servers.
Check that your corporate firewall, network, browser, or endpoint security controls are not blocking outbound traffic to the URL.
Why does authentication fail or not complete?
Sign in again with your Druva Console credentials.
If authentication still fails, check whether:
Your Druva session expired.
The browser sign-in flow was blocked.
The MCP-compatible agent blocked or did not complete the sign-in flow.
Your network or endpoint security controls blocked the authentication flow.
Why does a request return no data?
Verify that:
Your active Druva role has permission to view the requested workload, object, or target region.
Ensure that the target workload or time range you are asking about actually contains active data in your console.
Why does a request fail with a permission error?
Verify that your Druva role allows the requested data.
Existing Druva role-based access control (RBAC), permissions, and access boundaries apply when you use the Druva MCP server.
Why doesn’t the MCP-compatible agent ask for approval?
Do not assume that approval prompts are guaranteed.
Approval and confirmation behavior depends on the MCP-compatible agent. Existing Druva RBAC, permissions, and security controls still apply.
Why is my query not using the Druva MCP server?
AI agents can sometimes misroute generic prompts. Try being explicit: "Using the Druva MCP server, list the recent critical alerts."