N-able MCP

Overview, Tools, and Client Integration Guide

Connect to N-able MCP in 2 Simple Steps

N-able MCP lets you connect such AI tools as Claude, GitHub Copilot, Gemini and many more - directly to your N-able environment. Built on the N-able GraphQL API, it gives AI assistants real-time access to N-central device data, N-sight RMM data, and more via the N-able GraphQL API.

Ask questions in natural language, run compliance checks, and query your fleet without writing code or navigating separate product APIs.

View the full list of default and custom tools currently available.


1. Choose your endpoint

N-able MCP provides three distinct endpoints for different purposes. Choose the endpoint that matches your use case before configuring your AI tool or IDE.

Live data access and modification
Both `/mcp` and `/mcp/preview` can read and modify live production data across all connected N-able products. This includes executing scripts on devices and will extend to billing and other data as the platform grows. Understand the implications before connecting to a production endpoint. Use with caution - changes affect real customer environments.

Endpoint

Data access

Use case

https://api.n-able.com/mcp

Full read and write access across all connected N-able products.
Can query data, execute scripts, and modify devices.

Use this for production automations once you have fully tested your setup.
Treat it with the same care as direct access to your N-able platform.

https://api.n-able.com/mcp/read-only

Read-only. Queries and reporting only - cannot execute scripts or modify anything across any connected product.

Recommended starting point for most use cases.
Use this for dashboards, compliance reports, and any AI tool that needs visibility without write access.

https://api.n-able.com/mcp/preview

Preview. Early access to new tools and data sources before they reach production API. Also has full read and write access across connected products.

Use this to try new capabilities as they are developed.
Not recommended for production automations - tools and available data may change without notice.

If you open the MCP endpoint in a web browser you may see an RBAC: access denied message. This is expected because N-able MCP only accepts POST requests.


2. Connect your AI Tool or IDE (Non-Exhaustive Examples)

Before you start: Get your API Token

To connect N-able MCP to an AI tool or IDE, you need a valid API token. Tokens are tied to the N-able SSO user account that creates them, so the token only returns data that account has permission to see.

Get your API Token

Create your token at https://n-able.app/api-token-management. Copy it somewhere safe because you will not be able to view it again after creation.

Get your token

To communicate with N-able MCP, clients must provide a valid authorization token via the HTTP header:

Authorization: Bearer <YOUR_API_TOKEN>
Simpler setup is coming
We are working on dynamic client registration, which is planned for Q3 2026. This will let supported AI tools connect to N-able MCP without manually configuring a token. For now, use the token-based setup below.

Permissions

API tokens inherit permissions from the N-able SSO user who created them, so authorization depends on that user account's access.

If a token does not have adequate permissions:

  • Queries may fail with a 401 Unauthorized error
  • Fields your user account cannot access may return null rather than an error
  • Device lists will only include customers your account has permission to see

Example: if your token was created by a user with access to 3 of your 10 customer accounts, a query for all devices will only return devices from those 3 accounts.

N-able MCP works with any MCP-compatible AI tool or IDE. Find your tool below, copy the configuration, and restart the application. If your tool is not listed, the same pattern applies: use the server URL and pass your API token in the Authorization header.


Pick your tool

Claude Desktop

Claude Desktop is Anthropic's desktop-based AI assistant. Use this if you are using Claude as your AI tool for desktop.

Example:

{
"mcpServers": {
  "n-able": {
    "command": "npx",
    "args": [
      "-y",
      "mcp-remote",
      "https://api.n-able.com/mcp/mcp-read-only",
      "--header",
      "Authorization: Bearer <YOUR_API_TOKEN>"
  			]
  	}
  }
}

Claude Code

Claude Code is Anthropic's terminal-based AI assistant. Use this if you are using Claude as your AI tool from the command line.

Example:

{
  "mcpServers": {
    "n-able": {
      "type": "http",
      "url": "https://api.n-able.com/mcp/mcp-read-only",
      "headers": {
        "Authorization": "Bearer <YOUR_API_TOKEN>"
      }
    }
  }
}

Gemini CLI/Workspace

Gemini CLI lets you run AI queries from the terminal. Use this if you work with Google Workspace or prefer a command-line workflow.

Example:

{
  "mcpServers": {
    "n-able": {
      "httpUrl": "https://api.n-able.com/mcp/mcp-read-only",
      "headers": {
        "Authorization": "Bearer <YOUR_API_TOKEN>"
      }
    }
  }
}

GitHub Copilot - VS Code

GitHub Copilot is Microsoft's AI assistant built into Visual Studio Code. If you use VS Code as your editor, this is the integration to use.

{
  "servers": {
    "n-able": {
    	"url": "https://api.n-able.com/mcp/mcp-read-only",
    	"type": "http",
    	"headers": {
    		"Authorization": "Bearer <YOUR_API_TOKEN>"
    	}
    }
  },
  "inputs": []
}

After restart, an agent can invoke N-able MCP tools.

Choose your endpoint
Replace the URL in each config block below with the endpoint that matches your use case (from the Endpoints section above). The examples use `/mcp/preview` for safety, but you can swap in `/mcp` for production or `/mcp/preview`.

5. Verify it's working (Expected Behavior)

Once N-able MCP is configured properly:

✔️ There should be no console errors

  • No connection failures
  • No token-related errors
  • N-able MCP tools visible to the agent

N-able MCP server visible to the agent ✔ Requests authenticated using an API token ✔ Agent can resolve the schema and execute queries


Step 1: Quick connection check

After verification, you can test the connection by asking:

User input: How many devices are you managing?
"You are currently managing 42 devices across all your customers."
# Query may differ between AI models

query GetAssetCount {
  assetSearch(first: 1) {
    totalCount
  }
}
{
  "data": {
    "assetSearch": {
      "totalCount": 42
    }
  }
}

The agent should have used the default MCP tools to run a GraphQL operation and return an expected output like:

The number will reflect the actual device count in your tenant.

This proves that:

  • Authentication is correct
  • Default tools are available
  • The N-able GraphQL API is reachable

Step 2: Real MSP example

After the quick connection check, try a query that demonstrates an operational use case for MSP automation:

User input: Show me all Windows devices that don't have full BitLocker encryption, grouped by customer.
  "You have 28 Windows devices without full BitLocker encryption across your customers. Acme Ltd has the most unprotected devices. Would you like me to list them all or break it down by customer?"
# Query may differ between AI models

query UnencryptedWindowsDevices {
  assetSearch(
    first: 50
    where: {
      bitlocker: { protectionStatus: { notEquals: FULL } }
      operatingSystem: { type: { equals: WINDOWS } }
    }
    orderBy: [{ field: CUSTOMER_NAME, direction: ASC }]
  ) {
    totalCount
    nodes {
      name
      bitlocker {
        protectionStatus
      }
      customer {
        name
      }
      operatingSystem {
        name
        version
      }
    }
  }
}
 {
  "data": {
    "assetSearch": {
      "totalCount": 28,
      "nodes": [
        {
          "name": "DESKTOP-04",
          "bitlocker": {
            "protectionStatus": "NONE"
          },
          "customer": {
            "name": "Acme Ltd"
          },
          "operatingSystem": {
            "name": "Microsoft Windows 11 Pro",
            "version": "10.0.22621"
          }
        },
        {
          "name": "LAPTOP-12",
          "bitlocker": {
            "protectionStatus": "PARTIAL"
          },
          "customer": {
            "name": "Acme Ltd"
          },
          "operatingSystem": {
            "name": "Microsoft Windows 11 Pro",
            "version": "10.0.22621"
          }
        }
      ]
    }
  }
}

Current tools

View the full list of default and custom tools currently available.


Error Handling & Troubleshooting

ProblemLikely causeFix
401 UnauthorizedToken is missing, malformed, or has been revoked.Generate a new token at https://n-able.app/api-token-management and update your AI tool or IDE configuration.
404 Not FoundThe endpoint URL is wrong.Confirm the URL is https://api.n-able.com/mcp/preview with no trailing slash.
N-able MCP tools not visible in your AI toolThe client is misconfigured or was not restarted after setup.Re-check your configuration file, save it, and fully restart the IDE or AI tool.
Query returns no results or fewer devices than expectedThe token was created by a user who does not have access to all customers in N-able.Create a new token using an account with the appropriate customer access. See the Permissions section above.
execute returns an error but other tools workexecute is not enabled in your MCP client configuration.Check your MCP configuration and ensure execute is explicitly permitted. It is the only tool that requires this.
Security fields such as BitLocker return nullBitLocker data is only available for Windows devices. Non-Windows devices and some server editions will return null.Filter your query to Windows assets using operatingSystem: { type: { equals: WINDOWS } } before checking BitLocker status.
Your AI assistant reports an unknown fieldThe field name may have changed in a recent schema update, or the field does not exist.Use the search tool to find the correct field name, or check the schema changelog in the references below for recent changes.

Additional References

Share your feedback

N-able MCP is in active development and the tools added next will be shaped by what MSP automation workflows actually need. If something is missing, broken, or could work better, let us know.