mcp-ninjaone
# mcp-ninjaone
A [Model Context Protocol (MCP)](https://modelcontextprotocol.io) server for **NinjaOne** (formerly NinjaRMM), providing AI assistants with full access to remote monitoring and management through the NinjaOne REST API v2.
## Features
**19 tools** across five categories:
### Device Management
| Tool | Description |
|------|-------------|
| `list_devices` | List all managed devices with online status, OS, organization, and last contact |
| `get_device` | Get detailed device info including hardware, agent version, and network details |
| `get_device_activities` | Get recent activity log for a device (alerts, scripts, status changes) |
| `get_device_os_patch_status` | View pending, installed, and failed OS patches for a device |
| `get_device_software` | List all installed software on a device |
| `get_device_processor_info` | Get CPU/processor details for a device |
| `get_device_disks` | Get disk/storage info including capacity and free space |
### Organizations
| Tool | Description |
|------|-------------|
| `list_organizations` | List all organizations (clients/customers) with ID, name, and node count |
| `get_organization` | Get detailed organization info including contacts, tags, and policy assignments |
| `list_organization_devices` | List all devices belonging to a specific organization |
| `get_organization_locations` | Get configured locations for an organization |
| `get_organization_policies` | Get policy assignments for an organization |
### Alerts & Conditions
| Tool | Description |
|------|-------------|
| `list_alerts` | List active alerts across all devices, filterable by severity |
| `list_device_alerts` | List all active alerts for a specific device |
| `reset_alert` | Reset (acknowledge and clear) an active alert by its UID |
| `list_alert_conditions` | List all configured alert trigger conditions |
### Maintenance Windows
| Tool | Description |
|------|-------------|
| `list_maintenance_windows` | List all scheduled maintenance windows |
| `create_maintenance_window` | Create a maintenance window for a device to suppress alerts |
| `cancel_device_maintenance` | Cancel an active maintenance window to re-enable alerting |
### Jobs & Automation
| Tool | Description |
|------|-------------|
| `list_running_jobs` | List currently running jobs across all devices |
| `get_device_jobs` | Get all jobs (running, completed, failed) for a specific device |
| `run_script_on_device` | Execute a saved script on a device |
## Installation
```bash
git clone git@github.com:fredriksknese/mcp-ninjaone.git
cd mcp-ninjaone
npm install
npm run build
```
## Configuration
The server is configured via environment variables:
| Variable | Required | Description |
|----------|----------|-------------|
| `NINJAONE_CLIENT_ID` | Yes | OAuth2 client ID from NinjaOne API credentials |
| `NINJAONE_CLIENT_SECRET` | Yes | OAuth2 client secret from NinjaOne API credentials |
| `NINJAONE_INSTANCE` | Yes | NinjaOne instance hostname (see table below) |
### Instance Hostnames
| Region | Hostname |
|--------|----------|
| US | `app.ninjarmm.com` |
| EU | `eu.ninjarmm.com` |
| Oceania | `oc.ninjarmm.com` |
| Canada | `ca.ninjarmm.com` |
### Creating API Credentials
1. Log in to your NinjaOne instance
2. Navigate to **Administration > Apps > API**
3. Click **Add** to create a new API application
4. Select **Client Credentials** as the authorization flow
5. Grant the required scopes: `monitoring`, `management`, `control`, `offline_access`
6. Copy the **Client ID** and **Client Secret**
## Usage with Claude Desktop
Add to your `claude_desktop_config.json`:
```json
{
"mcpServers": {
"ninjaone": {
"command": "node",
"args": ["/absolute/path/to/mcp-ninjaone/dist/index.js"],
"env": {
"NINJAONE_CLIENT_ID": "your-client-id",
"NINJAONE_CLIENT_SECRET": "your-client-secret",
"NINJAONE_INSTANCE": "app.ninjarmm.com"
}
}
}
}
```
## Usage with Claude Code
Add to your Claude Code MCP settings:
```bash
claude mcp add ninjaone -- node /absolute/path/to/mcp-ninjaone/dist/index.js
```
Set the required environment variables before running, or configure them in your MCP settings.
## Example Prompts
Once connected, you can ask your AI assistant things like:
- *"Show me all devices that are currently offline"*
- *"What active alerts do we have across all organizations?"*
- *"Put device 1234 into maintenance mode for the next 2 hours"*
- *"List all devices for the Acme Corp organization"*
- *"What software is installed on device 5678?"*
- *"Show me recent activity for device 1234"*
- *"Run the cleanup script on device 9012"*
- *"What jobs are currently running across all devices?"*
- *"Reset all minor alerts for device 1234"*
- *"What patches are pending on device 5678?"*
## Development
```bash
npm run dev # Run with tsx (auto-reloads)
npm run build # Compile TypeScript to dist/
npm start # Run compiled output
```
## Architecture
```
src/
├── index.ts # Entry point — creates MCP server + STDIO transport
├── ninjaone-client.ts # HTTP client with OAuth2 client credentials flow
└── tools/
├── devices.ts # Device management and inventory (7 tools)
├── organizations.ts # Organization/client management (5 tools)
├── alerts.ts # Alert monitoring and acknowledgement (4 tools)
├── maintenance.ts # Maintenance window management (3 tools)
└── jobs.ts # Job execution and monitoring (3 tools)
```
The client handles OAuth2 token lifecycle automatically — tokens are fetched on first use and refreshed when they expire (typically after 1 hour).
## Requirements
- Node.js 18+
- NinjaOne account with API access enabled
## License
SEE LICENSE IN LICENSE
TDQS
Scored across 22 tools
Each tool targets a distinct resource and action. Tools like list_alerts and list_device_alerts are clearly differentiated by scope (global vs. per-device). No overlapping purposes.
All tools follow a consistent verb_noun pattern (e.g., get_device, list_organizations, create_maintenance_window). No mixing of styles like camelCase or inconsistent verbs.
22 tools is slightly above the typical well-scoped range (3-15) but justified given the breadth of RMM functionality covered (devices, alerts, maintenance, organizations, scripts).
Covers many key operations for device and alert management, but lacks full CRUD for devices, organizations, and scripts (e.g., no create_device, update_organization, or delete_script). Notable gaps in lifecycle coverage.