Security
Control is part of the product, not an afterthought.
Tessaire combines account safeguards, tenant-aware data controls, guarded infrastructure, and human review to protect customer catalog work. This page describes current practices without presenting them as a certification.
In brief
Security controls span identity, tenant isolation, secret handling, software delivery, and the review boundary around AI-assisted output.
Last updated 7 August 2026
Identity and account access
Every user signs in with a username and password and must complete TOTP multi-factor authentication. Recovery codes are single-purpose account-recovery tools; no SMS fallback weakens the second factor.
Passwords and recovery codes are stored as hashes. TOTP secrets are encrypted at rest with AES-256-GCM and protected by key material held outside source control.
Tenant and surface isolation
Tenant-owned records carry tenant identity through the data model. Tenant-leading composite keys and scoped queries prevent one workspace from being joined to another, while PostgreSQL row-level security provides a second enforcement layer.
Client and Tessaire HQ routes use separate application surfaces and access checks. Internal cost, margin, ROI, and operational records are not exposed through a client session.
Credentials and cloud access
Sensitive credentials are protected at rest using hashing or encryption appropriate to their purpose. Production secrets are stored outside the repository and are not placed in browser-delivered code.
Cloud delivery uses keyless workload identity rather than long-lived service-account key files. Deployment access is bounded to the Tessaire project and an explicitly selected release.
Software delivery safeguards
Changes pass local type, lint, test, documentation, and secret-scanning checks before integration. GitHub verification provides an additional public witness, while production deployment remains an explicit human-triggered action.
Secrets are scanned in the delivery path, and repository rules prohibit committing real client data or live credentials.
AI output remains governed
AI enrichment creates suggestions inside an asynchronous, auditable workflow. Provider identity and run details are recorded, and customer teams review the proposed values before approval.
A model response does not publish or overwrite approved catalog truth by itself. Human review is the control point.
Development data
Development and browser verification use fabricated demo catalogs and accounts. Project rules prohibit real client data in source code, fixtures, prompts, screenshots, and local test artifacts.
Responsible disclosure
Send a suspected vulnerability to Support@tessaire.net with the subject “Tessaire security report”. Include the affected area, steps to reproduce, and potential impact without accessing another customer’s data.
We will acknowledge and assess good-faith reports. Do not use denial-of-service testing, social engineering, destructive testing, or public disclosure before we have had a reasonable opportunity to investigate.
Tessaire does not claim SOC 2, ISO, or another security certification on the basis of these practices.
Questions
Ask for a clear answer.
One staffed address handles policy, accessibility, data, and security questions. Choose a specific subject so it reaches the right review.