Onam
Trust Center

Security at Onam

Onam reads your cloud — its resources and configuration — so it can tell you what you run and what is exposed. That makes our own security part of the product. This page explains, plainly, how Onam connects to your clouds, what each product reads and stores, and what we have and have not done yet.

Last updated 5 October 2026.

Compliance mapping

Onam’s product maps your findings to 78 compliance frameworks. That is a product feature, not a statement about Onam’s own certification.

How Onam connects to your clouds

You grant access with a template we provide, run in your own account, so you can read every permission before you deploy it.

CloudWhat you createAccess level
AWSAn IAM role that Onam assumes with an External IDAWS-managed SecurityAudit and ReadOnlyAccess policies, read of AWS Organizations account lists, and permission to start Onam's scan workflow in your account
AzureRole assignments for Onam's service principalBuilt-in Reader and Storage Blob Data Reader
Google CloudA service accountViewer, Cloud Asset Viewer and Security Reviewer; billing read only if you opt in
Oracle Cloud (OCI)A dedicated user and groupread policies across the tenancy, including users, groups, policies, vaults, keys and secret metadata
Alibaba CloudA dedicated RAM userA dedicated read-only RAM policy
IBM CloudA service IDViewer-level (read-only) access, scoped to a resource group

Posture scanning is read-only. It uses read permissions only.

Agentless workload scanning runs inside your account. If you enable it (AWS, Azure, Google Cloud), the template also creates resources in your account: a scan workflow, short-lived scan machines built from disk snapshots, and a storage bucket for results. Those resources hold the permissions needed to create and delete snapshots and scan machines; Onam’s own role can only start that workflow, not create or delete resources itself. Leftover scan resources are removed automatically after a set number of hours.

What Onam reads. AWS’s ReadOnlyAccess policy and Azure’s Storage Blob Data Reader role are broad enough to read stored objects, not only configuration. Onam requests them so data security posture management can locate where sensitive data lives. Posture scanning and data classification read configuration and metadata; they do not read the contents of your files, objects or database rows.

What each product reads and stores

Onam Security and Onam AI run in the same console, behind the same login, and work from the same connection to your clouds. Onam AI is enabled per organisation in early access. Here is what each reads, what it keeps, and what it does not do.

Reads
Cloud configuration and metadata through the access described above. Posture scanning and data classification do not read the contents of your files, objects or database rows.
Stores
Configuration and metadata from your clouds, findings, and the asset and attack-path graph — described in detail in the sections below.
Never
Posture scanning uses read permissions only. Agentless workload scanning, if you enable it, runs inside your own account.
Reads
Your Onam Security data, once Onam enables early access for your organisation. All inference goes through one model gateway — today Amazon Bedrock in the region agreed with you, one customer's data per call.
Stores
An append-only, hash-chained audit trail of every turn, tool call, decision and approval, and the evidence behind each claim. Sensitive values are masked in its answers.
Never
Changes nothing in your cloud without a person's approval; executing changes is not offered in early access. Customer data is never used to train or fine-tune a model.

What we store, and where

  • Hosting: Onam runs on AWS, in the region agreed with you — US, Europe, India or other regions — so you can meet your own compliance requirements.
  • What we store: configuration and metadata from your clouds, findings, the asset and attack-path graph, and your users’ account details, in PostgreSQL and a graph database (Neo4j).
  • Your cloud credentials: where a cloud needs a stored credential (for example an OCI API key or an Alibaba Cloud access key), it is kept in AWS Secrets Manager, which encrypts it with AWS KMS. For AWS we store no secret — Onam assumes your role.
  • Your source code (AI Code Fix only): when you run a fix, the repository is cloned for that run and the clone is deleted when it finishes. The Git token is used for that request only and is never stored or logged. See AI Code Fix.
  • Retention and deletion: customer data is kept for 30 days after a customer deactivates, then deleted.

Encryption

  • At rest: customer data is encrypted at rest. Stored credentials are encrypted by AWS KMS through Secrets Manager.
  • In transit: data is encrypted in transit for all customer-facing and service-to-service traffic, with one internal job being moved to TLS. Browser-to-Onam traffic uses HTTPS.

Signing in and access control

  • Single sign-on: SAML 2.0 (set up per organisation), OpenID Connect, Google and Microsoft sign-in.
  • Multi-factor authentication: use your identity provider’s MFA through single sign-on. Built-in MFA for password sign-in is not available yet.
  • Session cookies are HttpOnly, Secure in production, and SameSite=Lax.
  • Roles: users are invited into an organisation and can be limited to specific cloud accounts. Read-only roles cannot start scans or fixes.

Keeping customers apart

Every customer’s data carries a tenant identifier. In our main databases, PostgreSQL row-level security policies make the database itself refuse to return another tenant’s rows.

Logging and backups

  • Reads of the product’s data views are written to an audit log: who, what, when, from where, and the result.
  • Databases are backed up with AWS Backup.

Sub-processors

The third parties that process customer data on our behalf today:

Sub-processorPurposeLocation
Amazon Web ServicesHosting, databases, key management, and the model behind the AI assistant (Amazon Bedrock)The region agreed with each customer — US, Europe, India or other regions — to meet your compliance requirements
Mistral AIAI Code Fix only, and only when you run it: the content of each source file that has findings is sent to the model to generate the fixMistral AI's hosted API
Google WorkspaceEmail and collaborationGoogle's standard hosting regions

We will update this list before adding a new sub-processor.

Vulnerability disclosure

Found a security issue in Onam? Email security@onamsecurity.com. Please give us enough detail to reproduce it, and a reasonable time to fix it before telling others.

Safe harbor. We will not take legal action against research done in good faith that avoids privacy violations, data destruction and service disruption, and that tests only accounts you own.

Our machine-readable contact file is at /.well-known/security.txt.

Acknowledgments

We thank the researchers who report issues to us. None have been reported yet.

Ask us

Security questionnaire, DPA request, or a question this page does not answer? Write to us. We aim to answer standard security questionnaires within 2 working days.