# Security & Compliance — Security controls that pass audits and hold up in production.

> Enterprise deals shouldn't stall on security reviews, and preparing for SOC 2 shouldn't bury your engineers in manual busywork. We build practical access controls, private networking, secret management, and auditable change trails directly into your cloud accounts—producing real evidence your compliance team and auditors can rely on.

Canonical URL: https://cloudkite.io/security-compliance/

## How Daily Engineering Produces Audit Evidence

| Running in your cloud | Ready for your auditor |
| --- | --- |
| **SSO & least-privilege IAM** | Role mappings & access review logs |
| **Peer-reviewed IaC & deploys** | PR approvals & deployment history |
| **Automated backups & alerts** | Restore drill records & runbooks |

## Identity and access

Strong security starts with knowing who—and what—can reach production. We integrate directly with **Google Workspace / Cloud Identity, Microsoft Entra ID, and Okta** so permissions follow your company directory rather than scattered local accounts.

- **Least-privilege roles:** Separate permissions for developers, operators, and automated services across cloud IAM and Kubernetes RBAC.
- **Short-lived workload credentials:** Workload identity federation instead of long-lived static API keys that can leak.
- **Private administrative access:** Identity-aware network access through tools like **Tailscale**, keeping databases, clusters, and internal tools off the public internet.
- **Fast, repeatable onboarding and offboarding:** Group-based access that gets new hires productive in minutes and revokes access cleanly when roles change.

Your organization retains ownership of all accounts, identity providers, and infrastructure state; CloudKite access is explicitly scoped and removable whenever you choose.

### Keeping secrets out of code and containers

Application keys, database credentials, and API tokens belong in dedicated secret stores—never in Git repositories, CI logs, or container images. We wire your services and pipelines into the secret manager that fits your stack:

| Secret store | How we use it |
| --- | --- |
| **[Google Cloud Secret Manager](https://docs.cloud.google.com/secret-manager/docs/overview)** | Identity-scoped runtime secrets and automated rotation on Google Cloud. |
| **[AWS Secrets Manager](https://docs.aws.amazon.com/secretsmanager/latest/userguide/best-practices.html)** | Managed database credentials, rotation workflows, and IAM-scoped access on AWS. |
| **[Azure Key Vault](https://learn.microsoft.com/en-us/azure/key-vault/secrets/about-secrets)** | Centralized key, certificate, and secret access tied to Microsoft Entra identities. |
| **[1Password](https://www.1password.dev/secrets-automation)** | Shared team vaults and automated secret injection for engineering workflows. |

## Everyday operational security

Controls only protect the business if they still work on a busy release afternoon or during an incident:

- **Traceable infrastructure and code changes:** Pull-request reviews, Terraform plan history, container and dependency scanning, and a clear patching cadence.
- **Network isolation and encryption:** Private subnets, strict firewall boundaries, and encryption in transit and at rest tailored to the sensitivity of your data.
- **Prepared incident and recovery practices:** Centralized audit logging, high-signal alerting, automated backups, and tested restore runbooks so you know recovery works before you need it.

## Agent workload security

AI agents that execute code, call internal APIs, or read customer data introduce risks that traditional web apps don't. Following practical principles aligned with [OWASP's guidance on excessive agency](https://genai.owasp.org/llmrisk/llm062025-excessive-agency/), we enforce infrastructure-level guardrails outside the model itself:

- **Isolated sandboxes:** Dedicated execution environments, restricted filesystems, and hard CPU/memory budgets per workload.
- **Strict network egress:** Explicit allowlists controlling which model providers, internal services, and customer data stores an agent can reach.
- **Scoped authority and human gates:** Narrow, short-lived credentials enforced by the target system, with human approval steps around high-impact actions.
- **Full execution audit trails:** Structured logs and traces capturing tool calls, token usage, and access events for security and cost visibility.

## SOC 2 readiness and ongoing evidence

We work alongside your compliance lead and independent auditor to turn [SOC 2](https://www.aicpa-cima.com/resources/landing/system-and-organization-controls-soc-suite-of-services) requirements into working cloud controls—so audit season doesn't derail your engineering roadmap.

| Control area | Implementation and evidence we help maintain |
| --- | --- |
| **Access management** | SSO integration, group-based role assignments, periodic access reviews, and onboarding/offboarding records. |
| **Change management** | Required pull-request approvals, reviewed Terraform plans, and immutable deployment histories. |
| **Monitoring and response** | Centralized cloud audit logs, routed alerts, incident tracking, and practical response runbooks. |
| **System protection** | Managed secrets, encryption configuration, automated vulnerability scans, and tracked remediation. |
| **Recovery and continuity** | Scheduled backups, documented restore drills, and tested disaster-recovery procedures. |

## What this looks like in practice

When we partnered with the Miami Marlins to migrate off physical office servers onto Google Cloud, we replaced manual permission tickets with Google Workspace SSO, least-privilege IAM, and Tailscale access. New engineering hires went from slow manual setup to **productive access in under 30 minutes**, with cleaner audit visibility across every environment.

[Read the Miami Marlins case study](/marlins-case-study/).
