Block the change before it exists.
A CSPM tells you what is already broken. Ayra puts one gate in front of every change, from Terraform, CI, the console or the CLI, scores 170 guardrails in a single decision, and refuses the ones that should never ship. Every block carries a waiver path, and every verdict lands in a ledger your auditor can read.
One gate across every source.
A public bucket, an open SSH rule and an untagged GPU instance are the same kind of mistake: a change that should have been refused. Ayra scores security, cost and compliance guardrails together and returns one verdict, block, flag or allow.
- 170 built-in guardrails. Security, cost and compliance, curated from CIS, NIST, the Well-Architected frameworks and the FinOps Foundation, plus your own rules.
- Lifecycle governance on each one. An owner, a review date, a version and a rollout stage from draft to monitor to enforce. A guardrail that cannot be checked is listed, never counted as passing.
- Any source. Terraform plans, GitHub Actions, admission webhooks, the portal and the CLI all meet the same decision.
Native where the cloud allows it.
On Azure the gate publishes real Azure Policy deny rules, so the portal, the CLI, ARM and Terraform refuse the change at the control plane. On AWS, Ayra deploys Config managed rules where you enable them, which detect and can remediate. On Google Cloud and for pipelines everywhere, the gate runs as a CI check that fails the build.
- Azure: deny, today. Verified against a real tenant: the portal refuses provisioning with the guardrail's message.
- AWS: Config managed rules. Continuous evaluation per region, with CloudTrail polled for the change that caused the finding.
- Nothing blurred. Blocked at the gate and detected after the fact are labelled differently on every screen and in every export.
A waiver on every block.
A gate without a way through becomes a gate people route around. Each denial carries a signed break-glass link. The requester writes a reason, a security admin who is not the requester approves it for a fixed window, and the whole exchange lands in the ledger.
- Time-boxed and signed. Seven days by default, with the guardrail re-applied on expiry without anyone remembering to.
- Separation of duties in code. The approver is never the requester, and roles are scoped per product.
- Hash-chained, auditor-readable. Verdicts, waivers and remediations sealed with the previous hash and tagged to the control they evidence.
What is in Ayra.
The gate and the catalog
Block, flag or allow before provisioning. The full catalog across security, cost and compliance with owner, review date, version and rollout stage on every guardrail.
Exceptions
Waivers requested from the block itself, approved by a separate role for a fixed window, expiring on their own.
Drift and compliance mapping
Resources that left their approved state, and every guardrail mapped to ISO 27001, SOC 2, PCI-DSS, NIST 800-53 and CIS for your reporting.
Identity posture
Entra ID through Microsoft Graph: PIM gaps, over-privileged service principals, stale secrets, guest sprawl and conditional access drift.
AI and agent posture
Models, keys, filters and network exposure across Azure OpenAI, Bedrock and Vertex. Agents on Foundry, Bedrock and AgentCore, and Vertex AI Agent Engine, scored against OWASP Agentic and MCP controls.
Exploit intelligence
EPSS probability and the CISA KEV catalog weighed with blast radius and cost, so the few findings likely to be exploited on expensive resources come first.
Autopilot for security
Drafted fixes for misconfigurations, scored at the gate, approved by a security admin with typed confirmation, executed with a rollback hint.
Sensor bus
Defender for Cloud, Security Hub and Security Command Center findings arrive as sensors, deduplicated against the gate's own.
Control
Connections, security roles, webhook ingest for CI and admission payloads, channels for Slack, Teams, Jira and ServiceNow.
Mapped for your reporting, not claimed as ours.
Every guardrail carries the controls it evidences in ISO 27001, SOC 2, PCI-DSS, NIST 800-53 and CIS, and the mapping re-syncs nightly as the benchmarks change. Framework names describe coverage for your reports. The certification AIFog itself holds is listed on the Trust Center.
Your pipelines, your identity, your tools.
One outcome: a posture scan with findings recorded, and a gate your auditor can read.
Prevent Pro access for 90 days with execution withheld. The gate runs in monitor first so you see what would have been blocked, then your first enforce rule goes live with a waiver path.
Connect and scan
A subscription, account or project, or a CI webhook. The full posture scan records every finding under its guardrail, and lists what it could not check.
Monitor, then enforce one
Azure Policy in audit, AWS Config rules where enabled, CI checks reporting. Then one enforce rule, usually public storage or open SSH, with the first waiver landing in the ledger.
Show the auditor
Findings, verdicts and waivers in one hash-chained record mapped to the frameworks you answer to. Ayra pauses until you choose a plan; the record is kept.
Priced to fit your estate.
An environment is one Azure subscription, AWS account or Google Cloud project. Prevent includes three, Prevent Pro five, and we size the rest with you. There is no free rung; the free 90-day evaluation is the entry.
About Ayra.
How does the gate block a change before it exists?
What is a protected environment?
Who can approve a waiver?
Does Ayra also show cost guardrails?
What is covered under identity and AI posture?
How much does Ayra cost?
Start the proof of value
Connect one cloud or one Prevent source. The clock starts on the first, and the goal is a posture scan and a gate you can show an auditor.
Add Arian for the bill
Verified savings, allocation and Autopilot for cost. Both products on one organisation unlock the cost and risk graph, the unified gate and shared Autopilot, and we price them together.