Integrations
Integrations read data from hosting control panels, backup tools, cloud providers and Kubernetes through their APIs — read-only, with credentials stored encrypted and decrypted only for the request. They are under Integrations in the panel, for members whose role may manage servers. Each integration runs on its own interval, keeps the last answer and a history of answers, raises an alert when it starts failing, and turns stopped services, instances, pods and backup jobs into alerts.
Adding one
Integrations → Add integration, pick the type (grouped as Hosting panels, Backups, Cloud providers and Containers), fill in the fields and press Connect & check: the first check runs at once and the page tells you whether it worked. Common settings:
| Setting | Meaning |
|---|---|
| Name | your label |
| Check every | 5, 10, 15 or 30 minutes, 1, 6 or 24 hours (default 5 minutes) |
| Linked server (optional) | shows the panel data next to that server's agent metrics |
| Accept a self-signed certificate on the panel | for the types that connect to a host (panels, JetBackup); leave it off when the panel has a real certificate |
| Advanced → API base URL | most of the newer types have it; leave it empty to use the real endpoint. Set it only for a proxy or a test server |
Secret fields (tokens, passwords, keys) are never shown again; leaving one empty on edit keeps the stored value. An integration page has Run now, Disable / Enable and Delete (which removes its history too).
Every type described below as new in 2.1 shows its result the same way: a list of facts, tables (instances, sites, jobs, pods…), a services list (anything marked not OK becomes an alert), notes on what the provider's API cannot show, and warnings when only part of the data could be read (one region, one endpoint).
cPanel / WHM
What it shows: WHM version, hostname, load average (1/5/15), number of accounts and suspended accounts, disk used per account with the top 10, the service status list (running / enabled / monitored), partitions with usage, and the backup configuration (enabled, type, days, retention).
Setup:
- In WHM open Development → Manage API Tokens → Generate token.
- Give it a name and restrict it: tick only the privileges needed to read — List Accounts, View Server Information / View Server Status, Disk usage / Basic System Information, Backup configuration (names vary by WHM version). Do not grant Everything.
- Copy the token (shown once).
- In the panel: Add integration → cPanel / WHM, host (hostname or IP), port 2087, the WHM user (
rootor a reseller name) and the token. Tick Accept a self-signed certificate on the panel only if WHM still has a self-signed certificate.
A reseller token sees its own accounts; the account list may be missing for tokens without that privilege and the integration says so instead of failing. Requests go to https://host:2087/json-api/... with the header Authorization: whm user:token.
DirectAdmin
What it shows: version, number of users, services with on/off state, load average.
Setup: in DirectAdmin create a login key for the admin user (Admin → Login Keys) limited to the commands CMD_API_LICENSE, CMD_API_SHOW_ALL_USERS, CMD_API_SHOW_SERVICES, CMD_API_SYSTEM_INFO (or use the admin password, which is less safe). Add the integration with host, port 2222, the admin user and the key.
Plesk
What it shows: platform and version, hostname, number of domains and clients, services with running state.
Setup: in Plesk open Tools & Settings → API keys (or plesk bin secret_key --create with the Plesk REST API enabled) and create a key. Add the integration with host, port 8443, the admin user and the key. Requests use https://host:8443/api/v2 with X-API-Key.
CyberPanel (new in 2.1)
What it shows: connection check, CyberPanel version and build, websites with package and PHP version, the PHP versions in use.
Needs: host, port (8090), an administrator user name and password. In CyberPanel switch on API access for that administrator (Users → API Access). CyberPanel's API takes the admin user name and password in each request.
CyberPanel's API has no call for server load or service status: install the server agent for CPU, memory, disks and services.
Webmin / Virtualmin (new in 2.1)
What it shows: load average, CPU cores, memory, swap and disk use, the server status list (a stopped service becomes an alert) and the virtual servers with how many are disabled.
Needs: host, port (10000), the master administrator (root) and its password. Virtualmin's remote API (/virtual-server/remote.cgi, programs info and list-domains) answers only the master administrator, so there is no smaller permission to give.
aaPanel (new in 2.1)
What it shows: CPU cores and use, memory, load average, network rate and totals, partitions, the sites with their state, path and PHP version, and the installed runtime software (web server, database, PHP…) with whether it runs; stopped software becomes an alert.
Needs: host, port (7800) and the API key. In aaPanel open Settings → API, switch the API on, copy the key and add this monitor's IP address to the API whitelist. Only read calls are made (GetSystemTotal, GetDiskInfo, GetNetWork, the sites list and get_soft_list).
CloudPanel (new in 2.1)
CloudPanel has no public HTTP API (its clpctl tool is command line only), so this connection only checks what can be seen from outside:
- the admin panel answers over HTTPS on its port;
- the panel's TLS certificate: valid until, days left, issuer and subject;
- the CloudPanel version when the login page shows it.
Needs: host, port (8443) and optionally Warn when the certificate has fewer days left (default 14): fewer days raise an alert. No credentials. Install the server agent for CPU, memory, disks, services and sites.
JetBackup 5 on cPanel/WHM (new in 2.1)
A separate type (under Backups) so it can run on its own interval and alert on its own.
What it shows: backup jobs with type, destination, last and next run and state, the destinations and the backup queue.
Alerts: an enabled job whose last run failed, partially completed, was aborted or never finished, and an enabled job that is not running and whose next run is more than Overdue after this many hours late (default 2) in the past.
Needs: WHM host, port (2087), WHM user and a WHM API token (WHM → Development → Manage API Tokens). The same token as the cPanel / WHM integration works; JetBackup needs the root user or a reseller with JetBackup access. Only list calls are used (listBackupJobs, listDestinations, listQueueGroups). Add both the cPanel / WHM and the JetBackup integration to see the server and its backups. The agent can also read JetBackup locally, see Servers.
Cloudflare
What it shows: your zones (status, paused, plan, nameservers), per zone the last 24 hours of requests, bytes, threats, cached requests and unique visitors, and Cloudflare's own service status from cloudflarestatus.com.
Setup:
- In the Cloudflare dashboard open My Profile → API Tokens → Create Token → Custom token.
- Permissions: Zone → Zone → Read and Zone → Analytics → Read. Zone resources: all zones, or the ones you want to see.
- Copy the token and add the integration with it. Optionally enter one zone name to limit the view; empty shows up to 50 zones.
The token is verified against /user/tokens/verify first, so a wrong or expired token is reported clearly.
Cloud providers (new in 2.1)
Every cloud type lists the instances with their state and raises an alert for an instance that is not running (or, on AWS, whose status check is impaired). Tick Do not alert on stopped instances (wording differs per type) when you stop machines on purpose: they still show in the table but do not alert.
Amazon Web Services (EC2)
What it shows: EC2 instances per region with state, availability zone, type, system and instance status checks, address, launch time and, optionally, CloudWatch CPU (5-minute averages over the last hour).
Needs: Access key ID and Secret access key of an IAM user, Regions (comma separated) (default us-east-1), and Read CPU from CloudWatch (on by default; CloudWatch charges a small fee per metric read).
Minimum permissions: ec2:DescribeInstances, ec2:DescribeInstanceStatus and cloudwatch:GetMetricData. The managed policies AmazonEC2ReadOnlyAccess and CloudWatchReadOnlyAccess cover them. Requests are signed with AWS Signature Version 4.
Google Cloud (Compute Engine)
What it shows: Compute Engine instances in every zone of a project with status, machine type and addresses.
Needs: a Service account key (JSON): IAM → Service accounts → Keys → Add key → JSON; paste the whole file. Project ID (optional): empty means the project the key belongs to.
Minimum permissions: give the service account the Compute Viewer role only. The key is used to get a compute.readonly access token.
Microsoft Azure (VMs)
What it shows: virtual machines per subscription with power state, size, location and resource group.
Needs: Directory (tenant) ID and Application (client) ID (Microsoft Entra ID → App registrations → your app → Overview), a Client secret (Certificates & secrets → New client secret; the Value, not the Secret ID), and optionally Subscription IDs (empty = every subscription the app can read).
Minimum permissions: the Reader role for the app registration on each subscription.
DigitalOcean
What it shows: account status and droplets with status, region, size and addresses.
Needs: an API token: Control panel → API → Generate New Token with the Read scope only.
Hetzner Cloud
What it shows: the servers of one Hetzner Cloud project with status, type, location and addresses, plus the locations in use.
Needs: an API token (project): Cloud Console → your project → Security → API tokens → Generate, permission Read. A token sees one project; add one integration per project.
Vultr
What it shows: account balance and instances with status, power state, plan, region and addresses.
Needs: the account's API key (Account → API → Personal Access Token). Allow this monitor's IP address in the key's access control list.
Kubernetes (new in 2.1)
What it shows: cluster version and platform, nodes with Ready state, pressure conditions and capacity, pods (running, with problems, container restarts) per namespace, and deployments (desired vs available replicas).
Alerts: a node that is not Ready or reports pressure; a pod that is Failed or Unknown, waits with CrashLoopBackOff, ImagePullBackOff, ErrImagePull, CreateContainerConfigError, InvalidImageName or OOMKilled, or is Pending or Running-but-not-ready for more than 5 minutes; a deployment with fewer available replicas than desired (deployments scaled to 0 are ignored).
Needs:
| Field | Meaning |
|---|---|
| API server URL | the server: line of your kubeconfig, e.g. https://203.0.113.10:6443 |
| Bearer token | a service-account token (see below) |
| Cluster CA certificate (PEM, optional) | the certificate-authority-data of your kubeconfig, base64-decoded. Empty = the public CA list |
| Skip TLS verification | only for a lab cluster with a self-signed certificate |
| Namespace (optional) | empty = all namespaces |
Minimum permissions: a service account bound to the built-in view ClusterRole:
kubectl create serviceaccount aim-monitor -n kube-system
kubectl create clusterrolebinding aim-monitor --clusterrole=view --serviceaccount=kube-system:aim-monitor
kubectl create token aim-monitor -n kube-system --duration=8760h
A Role in one namespace also works for pods and deployments when you fill in Namespace; nodes always need cluster-wide read. For a cluster reachable only from inside, use kubectl on an agent host instead (see Servers).
Alerts from integrations
Integration alerts are not sent to alert channels or webhooks; use the agent or a monitor when you need to be paged.
- A service, instance, node, pod, deployment, backup job or certificate reported as not OK becomes an
integration.service_downalert in the alert history with an in-app notification for the organisation's members (repeated at most once an hour per item). - An integration that was OK and now fails (wrong token, host unreachable) raises
integration.erroronce (alert history and in-app notification); the status on the Integrations page shows the error text.
Linking to a server
An integration can be linked to a server you monitor with the agent so its panel data is shown on that server's page next to the agent metrics.
Limits
Integrations only show what each provider's API exposes; they are not a replacement for the agent on the same host. See Known limitations. The framework takes new types as a file under includes/integrations/ with an integration_<type>_fields() and integration_<type>_run() function.
One click helps us improve these docs.

