Worried about giving CloudMonitor access to your cloud? Read this first.

A direct, technical answer to the question every security-minded buyer asks: what can you see in our environment, and what can't you?

This is the post we wish every prospect would read before signing anything — including ours. When you connect a cloud account, a Kubernetes cluster, or a SaaS platform to a third-party monitoring tool, you are giving that tool an authentication credential. The right question to ask is not "is this vendor trustworthy?" (everyone says yes) but what does the credential they asked for actually allow?

The short answer

CloudMonitor asks for read-only, metadata-only access by default. Every credential we ask you to create is scoped to operations like List, Get, Describe, and Read — never Create, Update, or Delete. There is no exception. Where we suggest a fix, we show you the command and you run it yourself — CloudMonitor makes no write call to your cloud.

What "metadata" means in practice

It means we read the shape of your environment, not its contents:

Why the IAM policies we ask for are public

Every IAM role / RBAC binding / token scope we recommend is in the Connectors guide. You can read the exact permissions before you grant them. You can also tighten them further — if you want to exclude an entire service from a scan, drop the corresponding permission from the policy. That one scanner will report an access error; the others carry on.

The data we do store

We store the results of scans — inventory snapshots, finding records, cost time series, activity-log entries — in our own database, with every row keyed to your tenant and every query scoped to it. We do not store your credentials in plaintext: a secretless Azure connection leaves us no credential to store, and the secrets, access keys and API tokens we do hold are encrypted at rest with AES-256-GCM under a key kept outside the database. That protects them if the database or a backup is compromised; it does not make them unreadable to the running application, which has to decrypt them to scan. The key is one application-level key, not a key per customer, and there is no bring-your-own-key option. Our Privacy Policy says the same thing.

Who on our side can see it

No other customer can see your data. The operator's own staff roles can: administrative access to the application, its database and its logs is used to run the service, diagnose faults and answer support requests. Section 9 of the Privacy Policy says exactly what that covers. Any vendor who tells you nobody on their side can read your scan results is either using per-customer keys they cannot access, or not telling you the truth; we are not the first kind.

The five questions you should ask any monitoring vendor

  1. What IAM policy do you ask for, exactly? If they can't paste it into chat, that's a red flag.
  2. Do you support credential-less auth? Managed Identities, Workload Identity Federation, IAM Roles for Service Accounts — these mean the vendor never sees a long-lived secret.
  3. Do you read data-plane contents or only control-plane metadata? Most legitimate monitoring tools only need the latter.
  4. Where is my data stored, and can I delete it? Tenant isolation, deletion timeline, audit log of admin access.
  5. What happens if I revoke the credential? Should degrade gracefully. Should never get re-granted automatically.

How we answer question 2 ourselves

Partly. On Azure there are two ways to connect. The secretless one uses workload identity federation: you add CloudMonitor's multi-tenant application to your directory, it trusts our scanner's managed identity, and we hold no secret for your tenant at all. It is in Beta, not yet validated against a live customer tenant. The way known to work today is a service principal with a client secret, which we store encrypted; certificate credentials are not supported yet. On AWS and GCP we are not there yet: today the AWS connector uses an access key for a dedicated IAM user with a metadata-only policy, and the GCP connector uses a service-account key. Both are long-lived secrets that we store encrypted. Cross-account roles and Workload Identity Federation are what we want to move to, and neither is built. The AWS and GCP connectors are also in Beta: verified against emulated APIs, not yet validated on a live customer account.

CloudMonitor holds no SOC 2, ISO 27001 or other certification or attestation, and offers no SLA. If you have a specific compliance question, write to us via contact. We'd rather answer the hard question on day one than after you've signed.

← Back to all posts