Main content

Security & compliance

Enterprise-grade
trust & security

The Varcio platform is built with security at every layer — from credential encryption to immutable audit trails and compliance control mapping.

Varcio platform · ComplianceLive workspace
Varcio Compliance dashboard showing 87.1% control coverage mapped to SOC 2, CIS, NIST and ISO controls, open SLA breaches, and mean time to resolution
SOC 2Control mapping
99.9%Uptime SLA
AES-256Encryption at rest
TLS 1.3Data in transit
0Plaintext secrets

Security by design, not afterthought

Four foundational principles guide every architectural and product decision.

Zero trust by default

No implicit access. Every API call is authenticated, authorized, and rate-limited.

No plaintext secrets

AWS uses AssumeRole, while Azure, GCP, and OCI credentials are encrypted at rest and never exposed in logs, API responses, or the UI.

Immutable audit trail

Every action, decision, and data access is logged with actor, timestamp, payload, and outcome. Logs cannot be modified or deleted.

Defense in depth

Multiple independent security layers — authentication, authorization, execution safety, rate limiting, and protected resources.

Four layers around every cost action

Credentials never sit in plaintext, permissions are scoped twice, and every write passes human approval or Autopilot’s confidence and budget limits before it reaches a resource.

Layered security model around every cost actionFour nested layers protect any write action: provider-native trust, encrypted credential storage, dual-layer role-based access control, and a required human approval gate at the centre.Provider-native trustAssumeRole · SP · SA JWTEncrypted credentialsAES-256 at rest · TLS 1.3Dual-layer RBACpage + action permissionsApproval gatehuman sign-off requiredProtected tags keep critical infrastructure out of scope entirely
Each layer narrows what an action can reach — with the human approval gate at the centre.

From finding to action, fully audited

Detection is automated. Action clears policy and a named approver first, unless you turn on Autopilot, which acts only above confidence thresholds and within a monthly budget.

Approval-gated remediation pipelineA finding moves through detection, scoring, policy evaluation, human approval in Slack or Teams, and finally execution — with every step written to an immutable audit log.Detect316 detectors1Scoresavings · risk2Policyadvisory / block3ApproveSlack or Teams4Executeprotected tags5HUMAN GATEImmutable audit log — actor, timestamp, and payload recorded at every step
Actor, timestamp, and payload are recorded at every step for SOC 2 evidence collection.

How your data moves through the platform

From connection to action — every step is encrypted, audited, and gated.

  1. Connect

    Provider-native trust models with no plaintext credential exposure

    AWS: AssumeRole with External ID. Azure: encrypted Service Principal. GCP: encrypted service account key with short-lived tokens. OCI: encrypted API signing key.

  2. Ingest

    Cost data pulled via official cloud APIs

    AWS Cost Explorer and Cost and Usage Report, Azure Cost Management, GCP BigQuery billing export and the OCI Usage API — read with read-only access.

  3. Process

    Data normalized, analyzed, and stored encrypted

    Cost metadata is encrypted at rest, encrypted in transit, and processed inside isolated application environments.

  4. Act

    Remediation runs behind approval and guardrails

    Requested changes need approval; Autopilot, if enabled, acts only above confidence thresholds and within a monthly action budget. Protected tags exclude critical resources, and rollout-safety gates can hold a run to dry run.

9 security control layers

Every layer of the platform is designed with defense-in-depth principles.

Access

Authentication

JWT sessions, OAuth (Google, GitHub, Azure AD), SSO via OIDC and SAML 2.0, SCIM provisioning (Enterprise)

Access

Authorization

Dual-layer RBAC: page permissions (navigation) + action permissions (operations)

Data

Credential storage

AWS KMS envelope encryption: each record gets its own AES-256 data key and the content is sealed with Fernet — zero plaintext persistence

Access

API tokens

SHA-256 hashed, show-once creation, configurable expiry, instant revocation

Runtime

Execution safety

Single-use JTI tokens, replay prevention, approval gates for all write operations

Integration

Webhook signing

HMAC-SHA256 per-subscription signing secrets

Runtime

Rate limiting

Per-IP and per-route limits with elevated thresholds for compute-heavy endpoints

Compliance

Audit trail

Immutable action and decision logging with actor, timestamp, and payload details

Compliance

Compliance

Findings mapped to controls in 8 frameworks, including SOC 2, CIS, PCI DSS, ISO 27001 and NIST 800-53, with evidence export

Encryption at every layer

LayerMethodScope
Data at RestAES-256 encryptionAll database fields and stored files
Data in TransitTLS 1.3All API calls, webhooks, and integrations
CredentialsAWS KMS envelope encryptionPer-record AES-256 data key; cloud connection credentials sealed with Fernet
API TokensSHA-256 hashUser and service API tokens
Webhook SignaturesHMAC-SHA256Per-subscription signing secrets

Provider-specific trust models

AWS, Azure, GCP, and OCI are handled differently, but every path avoids plaintext exposure and keeps execution behind approval gates.

ProviderMethodHandling
AWSIAM AssumeRoleNo static access keys required or stored
AzureEncrypted Service PrincipalStored encrypted, never exposed in logs or UI responses
GCPEncrypted service account keyShort-lived access tokens minted per call; never exposed in logs or API payloads
OCIEncrypted API signing keyPrivate key, fingerprint and optional passphrase stored encrypted, never exposed in logs or UI responses
ExecutionVerified write permissionsAWS uses a separate execution role with its own External ID; Azure, GCP and OCI write permissions on the same credential are checked separately. Approval or Autopilot limits apply before live execution

Dual-layer RBAC

Page-level visibility and action-level permissions — independently configurable per role.

RolePagesActionsDescription
ViewerDashboard, FindingsView-onlyRead-only access to cost data and findings
AnalystAll except AdminScan, Query, ForecastRun analyses but cannot execute changes
OperatorAll except AdminScan, Query, Approve, RemediateFull operational capabilities within team scope
AdminAll pagesAll actions + SettingsWorkspace management, API keys, integrations

Who can sign in, from where, and how

Access follows your directory, and automation gets its own identities instead of borrowing a person’s.

Single sign-on

OIDC and SAML 2.0 with signed assertion validation. Identity provider discovery by email domain routes each person to the right provider. Enterprise plan.

SCIM provisioning

Your identity provider creates, updates and deactivates accounts, with group-to-role mapping and a bearer token you can rotate or revoke. Enterprise plan.

Sign-in policy

Turn SSO on, require it for everyone, require two-factor authentication, and decide whether a first sign-in creates an account.

Two-factor for administrators

Customer administrator two-factor authentication is on by default in production, with a 14-day grace window. Sessions started through enterprise SSO are exempt because your identity provider owns that factor.

Network access

Limit sign-ins, sessions and API tokens to approved office or VPN networks.

Service accounts

Non-human identities for automation and Terraform. Roles decide what they can reach and each scoped API key decides what it can call. API tokens are hashed and shown once.

SOC 2 readiness framework

Findings mapped to eight SOC 2 Trust Services Criteria controls. Mapping supports evidence collection; it is not a certification.

CC6.1Mapped

Logical access controls

Findings on access restrictions that protect information and limit access to identified users

CC6.2Mapped

Credential lifecycle

Findings on unused, old or unrotated credentials and keys

CC6.3Mapped

Role-based access

Findings on roles and permissions broader than needed

CC6.6Mapped

Boundary protection

Findings on resources and network paths exposed outside your boundary

CC7.1Mapped

Vulnerability detection

Findings on detection and remediation of vulnerable configuration

CC7.2Mapped

Anomaly monitoring

Findings on the logging and monitoring needed to spot anomalies

CC8.1Mapped

Change management

Findings on controls over configuration change

A1.2Mapped

Recovery infrastructure

Findings on backups, deletion protection and recovery readiness

How we compare

The only option in this comparison that combines AI operations, commitment purchasing, policy enforcement, and governed remediation in one product.

Scroll sideways to see every column.

CapabilityCloudHealthInfracostKubecostVarcio
Multi-Cloud Spend——
Pre-Merge PR Analysis——
K8s Cost Attribution——
AI Conversational Interface———
Automated Remediationlimited——
Commitment Purchasing———
Predictive Forecastingbasic—basic
ChatOps Approvals———
Event Bus / Webhookslimited——
Policy Enginebasicbasic—
Starting Price$$$$Free (limited)Free (limited)$49/mo

Security FAQs

Does Varcio store our cloud credentials?

AWS uses IAM AssumeRole, so no static access keys are stored. Azure Service Principal secrets, Google Cloud service account keys and OCI API signing keys are stored with AWS KMS envelope encryption, where each record gets its own AES-256 data key and the content is sealed with Fernet. They are never exposed in logs, API responses, or the UI.

What data do you access?

Cost, usage and billing data; commitment coverage and utilisation; resource inventory and configuration; utilisation metrics; and recommendations from each provider’s own advisors. For diagnostics, Varcio can also query exported database logs, such as slow-query and deadlock logs, through CloudWatch Logs Insights, and read query performance for databases you connect to health monitoring. Access is limited to what the credentials you grant allow.

Can auto-remediation break production?

Several safety layers apply: protected-tag exclusion, rollout-safety gates (step-up approval, a rollback success floor and maintenance windows for larger blast radius, otherwise dry run), per-rule dry-run and action caps, a maximum number of actions per run, and a monthly action budget. Requested changes need human approval. In Autopilot mode, some actions can run without a person above a confidence threshold, but resources that look production, public, critical or deletion-protected go to approval instead. Resources tagged as protected are never touched.

Are you SOC 2 certified?

The Compliance module maps findings to controls in 8 frameworks: CIS AWS, CIS Azure, CIS Google Cloud, SOC 2, PCI DSS v4.0, ISO 27001:2022, NIST 800-53 Rev. 5 and Cloud Well-Architected, with evidence export. SOC 2 mappings cover CC6.1–CC6.3, CC6.6, CC7.1, CC7.2, CC8.1 and A1.2. Mapping findings to controls is not a certification. Formal certification is in progress.

How do you handle data residency?

Cost data is processed and stored in the region you choose during onboarding. Enterprise plans support custom data residency requirements.

What happens if you have a breach?

Because AWS uses AssumeRole and Azure, GCP and OCI credentials are envelope-encrypted with strict exposure controls, the blast radius is limited to the data and write permissions you have granted. Incident response, notification, and immutable audit trails support forensic review.

Secure. Compliant. Ready.

Start your free trial with enterprise-grade security from day one.