Integration inventory
Which controls need which systems, and which of those have a supported connector already.
Warden · Deployment
Your evidence lives in systems no vendor has a connector for
Engineering work to connect Odingard to your estate — cloud accounts, identity, source control and the internal systems that only you run — so control monitoring covers what actually matters.
Standard connectors cover the standard estate
AWS, GCP, GitHub and Cloudflare are supported out of the box. The controls that fail an audit are usually the ones monitored through an internal deployment system or a mainframe-adjacent process with no public API.
- The critical system is often the custom oneThe internal change management tool is exactly the control an auditor samples.
- Read-only access takes negotiationGetting appropriately scoped credentials across account boundaries is organisational work as much as technical.
- An unmaintained integration is a silent failureA connector that stops returning data looks identical to a control that is passing.
How the engagement runs
Access design
Least-privilege read-only credentials, scoped per account, with a rotation plan.
Build
Custom collectors for the systems with no supported connector, written against your interfaces.
Monitor the monitors
Staleness detection so a broken collector raises an alert rather than reporting a pass.
What you hold at the end
- Supported connectors configured across AWS, GCP, GitHub and Cloudflare as applicable
- Custom collectors for internal systems, with source and documentation handed over
- Least-privilege credential design with a rotation plan
- Staleness detection so a failed collector cannot masquerade as a passing control
- An integration runbook your platform team can maintain
What is integrated
Coverage where it counts
A monitoring program that only covers the systems with easy APIs will be sampled precisely where it does not.