Privacy policy.

How we use your data.

• —Operating the platform. Your infrastructure data powers the knowledge base, chat context, and script execution. This is the product.
• —Security and audit. IP addresses, user agents, and action logs are used for session management, anomaly detection, and audit trails.
• —Transactional email. We send emails related to demo requests and account operations. No marketing email lists. No newsletters unless you explicitly opt in.
• —Improving relevance across the platform. Aorka computes mathematical representations (vector embeddings) of knowledge items and groups them into concept clusters. These clusters determine which types of knowledge are most relevant across the platform — for example, recognizing that firewall-configuration facts are broadly important right now. This process operates on abstract coordinates, not on the content of your data. No titles, fact values, hostnames, or other readable information is shared between tenants. See Concept clusters below for the full mechanism.
• —We do not sell your data. Not to advertisers, not to data brokers, not to anyone. Your infrastructure data is yours.

Tenant isolation.

Aorka is a multi-tenant platform. Your data — endpoints, facts, conversations, credentials, scripts, job history — is scoped to your tenant at the database level. This is not application-layer filtering that could be bypassed; it is structural. No other tenant can see, query, or access your data through the application.

• —Role-based access within tenants. Within your tenant, access is controlled by role-based permissions (admin, tech, viewer) and unit-scoped access grants. A user only sees the endpoints and data they have been explicitly granted access to. Tenant admins control who gets access to what.

Concept clusters and cross-tenant signals.

While your data content is strictly tenant-isolated, Aorka uses a global relevance system to improve knowledge quality across the platform. Here is exactly how it works:

1. 1
   Vector embedding. Each knowledge item is converted into a vector embedding — a list of numbers representing the item's meaning in abstract mathematical space. This is a one-way transformation; the original text cannot be reconstructed from the vector.
2. 2
   Cluster centroid. Vectors from all tenants are grouped into concept clusters using density-based clustering. A concept is a geometric centroid (average position) of nearby vectors — it has no name, no label, and no readable content. It is a point in mathematical space.
3. 3
   Salience score. Each concept has a salience score reflecting how actively that type of knowledge is being used across the platform. When many items in a concept cluster are accessed frequently, the concept's salience rises; when activity is low, it falls.
4. 4
   Local propagation. Salience changes propagate to nearby items. If the “firewall configuration” region of the vector space is heating up across the platform, your firewall-related items benefit from that signal — they rank higher in search results and get validated more frequently.

What crosses tenant boundaries: mathematical coordinates and numerical relevance scores. What never crosses: titles, descriptions, fact values, hostnames, IP addresses, credentials, or any other readable content. No tenant can reverse-engineer another tenant's data from concept clusters — the vectors are abstract, high-dimensional, and irreversible.

Retention and deletion.

• —You control your data. Conversations, facts, credentials, and knowledge base entries can be deleted by authorized users directly through the platform. Deletion is immediate — we do not soft-delete and retain data behind the scenes.
• —Account deletion. Tenant admins can request full account deletion by contacting us. We will delete all tenant data — endpoints, facts, conversations, credentials, scripts, and user records. Audit logs may be retained for up to 90 days after account deletion for security purposes, after which they are permanently deleted.
• —Backups. AWS RDS automated backups are retained per AWS default retention policy. When data is deleted from the live database, it will naturally age out of backups as they rotate.

Changes to this policy.

If we make material changes to this privacy policy, we will notify active users via the platform and update the “Last revised” date at the top of this page. We will not retroactively reduce your privacy protections without notice.