Skip to main content

Overview

Multi-tenant applications need strong isolation between tenants while sharing common infrastructure. Bedrock Cloud provides built-in multi-tenancy with complete isolation between customers.

Bedrock Cloud Multi-tenant Architecture

Each tenant is completely isolated. Users, roles, and permissions in one tenant cannot access another tenant’s resources.

Setting Up Multi-tenancy

1. Create Tenants via Management API

2. Build Tenant Structure

3. Define Project-Level Roles

Roles and permissions are defined at the Project level:

Tenant Isolation

Users Can Only Access Their Tenant

Verifying Tenant Isolation

Agents in Multi-tenant Environments

Tenant-specific Agents

Each tenant can have their own AI agents scoped to their projects:

Agent Delegation

Agents should act on behalf of users for proper access control:

Permission Patterns

Project-Level Permissions

Environment-Specific Overrides

Querying Tenant Access

Determine which projects a user has access to:

Best Practices

Roles and permissions should be defined at the project level, not the tenant level. This allows each project to have its own authorization model.
Apply overrides to production environments to restrict dangerous operations.
Don’t give agents tenant-wide access. Scope them to the specific projects they need.
When agents act on behalf of users, use the onBehalfOf field to ensure proper access control.
Use externalId on tenants and users to map to your billing/CRM systems.

Next Steps

Scope Overrides

Fine-tune permissions at tenant level

Agent Governance

Manage agents across tenants