Documentation

Multi-Tenant Isolation

How Fabric AI enforces strict data isolation between personal accounts, organizations, and users.

Fabric AI is a multi-tenant platform where every feature supports both personal and organization contexts with strict data isolation. This page explains how your data is protected.

Isolation Boundaries

Data isolation is enforced at three levels:

BoundaryRule
Personal vs. OrganizationA user's personal data is never visible in an organization context, and vice versa
Organization vs. OrganizationOrganization A's data is never visible to Organization B, even if the same user belongs to both
User vs. User (within org)Some data is user-private within an organization, while other data is shared across members
Loading diagram...

How Your Data is Protected

Fabric enforces data isolation at multiple layers to ensure your information stays secure.

Application-Level Isolation

Every API request is scoped to the correct context. When you're working in a personal context, only your personal data is accessible. When you switch to an organization, only that organization's data is shown. There is no way to accidentally query across contexts.

Database-Level Security

In addition to application-level filtering, the database itself enforces row-level security policies. Even if an application bug were to bypass the normal filtering, the database would reject unauthorized access. This defense-in-depth approach ensures no single point of failure can compromise data isolation.

Vector Search Isolation

Document embeddings used for AI-powered search and retrieval are isolated per tenant. Searching for documents only returns results from your current context.

File Storage Isolation

Uploaded files (documents, avatars, exports) are stored with tenant-scoped access controls, preventing cross-tenant file access.


Personal vs Organization Context

Fabric provides two distinct contexts for organizing your work:

Personal Context (/app/...)

Your private workspace. Everything you create here is visible only to you:

  • Personal projects and documents
  • Personal AI chat conversations
  • Personal agent configurations
  • Personal integration credentials

Organization Context (/app/{org-slug}/...)

A shared workspace for your team. Data created here is scoped to the organization:

  • Organization projects visible to all members
  • Shared agent deployments
  • Organization-wide AI configurations
  • Team collaboration on documents and workflows

Use the context switcher in the top navigation to move between personal and organization contexts.


What's Isolated

Fully Isolated Per Tenant

These resources are strictly separated between personal and organization contexts:

  • Projects and project documents
  • AI Chat conversations and history
  • Agents — templates, deployments, and memory
  • Workflows and execution history
  • Workspaces and uploaded documents
  • Prompts and prompt templates
  • Reports and report executions
  • Purchases and billing

Per-User Within Organizations

Some resources are private to each user even within an organization:

  • Integration credentials — Each member manages their own API connections
  • AI provider keys — Personal API keys are never shared with the organization

Shared Within Organizations

  • System agents and system prompts are available to everyone
  • Organization-level AI settings apply to all members (unless overridden personally)

Organization Roles

Within an organization, roles control what members can do:

RolePermissions
OwnerFull control including billing, deletion, and member management
AdminManage members, settings, and all resources
MemberUse agents, create documents, and access shared resources

Continuous Verification

Fabric continuously tests tenant isolation to ensure data boundaries are maintained. Automated test suites verify that:

  • Personal data never appears in organization contexts
  • Organization A's data never appears in Organization B
  • Context switching correctly changes the visible data set
  • New features maintain isolation from the start

Next Steps