Trust & Security
How Sagaris protects your data — stated plainly, including what is done, what is in progress, and what is not in place yet. We would rather show you an honest roadmap than a wall of badges.
Hosting & data location
Sagaris runs on Google Cloud Platform (Cloud Run and Cloud SQL PostgreSQL) in the us-east4 region. Infrastructure is managed as code, and secrets live in GCP Secret Manager — never in source control or container images.
Encryption
TLS protects data in transit (HSTS enabled; database connections require TLS). Data at rest uses GCP-managed encryption, and high-value credentials such as OAuth tokens are additionally encrypted with AES-256-GCM at the application layer.
Access control
Authentication uses Google Identity Platform session cookies with company-email confirmation codes, verified with revocation checks that fail closed. Authorization is an eight-role RBAC model with least-privilege, fail-closed defaults, and every data access is scoped to your workspace.
Resilience
Automated database backups with point-in-time recovery and deletion protection are enabled. Background work runs on durable job queues, so no in-flight work depends on a single server instance. Known open item: the primary database is single-zone today; regional high availability is a tracked infrastructure change.
Secure development
Every change ships through reviewed pull requests with required code-owner review on security-sensitive paths, CI gates (build, lint, type-check, tests), automated dependency vulnerability scanning (Dependabot, npm audit, govulncheck), and fail-closed database migration governance.
Monitoring
Structured logs with automatic redaction of sensitive values, audit trails for exports, imports, and AI-initiated actions, plus uptime checks and alerting on availability, error bursts, and compliance violations.
These six product boundaries are commitments built into the platform, shown to every workspace owner in the product itself.
Does not send outreach without explicit human approval.
Drafts are generated for review. A person confirms each send (or an opt-in auto-send rule) before any message leaves the workspace.
Does not write back to your CRM unless write-back is enabled.
Sync is read-first. CRM write-back is opt-in per object and follows the conflict-resolution matrix below; Sagaris-owned fields never overwrite your reps blind.
Does not extract fields you place on the do-not-extract list.
An excluded field is never read into the model context, even when its parent object is included for extraction.
Does not train shared models on your workspace data.
Your records are used to serve your workspace only. They are not pooled into a cross-customer training set.
Does not delete CRM records on your behalf.
Sagaris creates and updates within the limits you set. Destructive deletes stay a human action in your CRM.
Does not expand its data scope without a configuration change.
What is extracted is governed by the toggle matrix on this page. Adding a new object or field to extraction is always an explicit, logged change.
SOC 2: internal control matrix complete, attestation pending
We completed a full SOC 2-mapped technical control audit of the platform in July 2026, with a maintained remediation gap register. We do not yet hold a SOC 2 report — engagement of a third-party auditor is a planned milestone — and we will not describe ourselves as certified until an independent auditor says so. Security reviewers can request the control matrix and gap register.
Penetration testing: not yet conducted
A third-party penetration test has not yet been performed. Internal security audits (application security, access control, secrets) have been completed and are available to reviewers on request. When an external test is completed, its summary will be published here.
Enterprise access: SCIM built and gated, SSO on the roadmap
SCIM 2.0 provisioning is built in the platform but gated and not yet generally available. SAML/OIDC single sign-on is on our roadmap and has not been built yet; today, sign-in uses Google Identity Platform with company-email confirmation codes. Contact us if either is a requirement for your evaluation.
- Consent-first outbound: per-contact, per-channel consent tracking with regional rules (GDPR, CAN-SPAM, TCPA, CCPA/CPRA, CASL, PECR) and non-overridable safeguards such as mandatory unsubscribe links and do-not-call screening.
- Right to erasure is implemented: personal data is anonymized, hashed suppression records prevent silent re-import, and a verifiable deletion certificate is produced.
- Your data is exportable: contact and workspace data can be exported in CSV form, and data-subject-request activity is tracked against statutory deadlines. A dedicated self-service request portal is on our roadmap.
- Opt-outs are honored immediately across all sequences for that channel.
We use a defined set of service providers to operate the platform. Several only process data when you enable the relevant feature or integration. A full list with purposes and data categories is available to security reviewers on request.
| Category | Providers |
|---|---|
| Cloud & AI infrastructure | Google Cloud Platform (hosting, database, auth), Vertex AI Gemini (primary LLM), OpenAI and Anthropic (optional fallback LLMs) |
| Communications | Telnyx (telephony/SMS), Twilio (legacy telephony paths), Resend (email delivery), ElevenLabs (voice synthesis) |
| Business & enrichment | Stripe (billing), Apollo.io (enrichment), Perplexity (research) |
| Customer-enabled integrations | Salesforce, HubSpot, Gmail/Google Calendar, Microsoft Outlook (Entra ID), LinkedIn, X |
| Domains & DNS | Cloudflare, GoDaddy, Porkbun |
We take vulnerability reports seriously and acknowledge them within 2 business days. Please report suspected vulnerabilities privately through GitHub Security Advisories rather than a public issue, per our published disclosure policy.
Read our security disclosure policy (SECURITY.md)