In a nutshell: A pool model for multi-tenancy on Bedrock AgentCore enables logical isolation with shared infrastructure through scoping, access policies, and data partitioning.
AWS demonstrates architectures for production-ready multi-tenant systems with Amazon Bedrock AgentCore that enable complete tenant isolation, service tiers, and granular cost allocation without dedicated infrastructure per tenant.
Amazon Bedrock AgentCore offers native mechanisms for implementing multi-tenant systems without requiring dedicated resources per tenant. The approach framework follows a three-level hierarchy of Tier → Tenant → User, with isolation at each level enforced through access controls on Knowledge Base, Memory, model access, and cost tracking.
The solution demonstrates a pool model in which tenants share the same infrastructure and compute resources, but logical separation is guaranteed through scoped identifiers, IAM policies, and data partitioning. A tiering strategy complements this: Basic-tier tenants work with the smaller, cost-efficient Mistral Ministral 3 8B Instruct and receive simple document search; Premium-tier tenants access GPT-4-equivalent models and web search tools. This differentiation occurs through model selection and tool access per tier.
Concretely, cost tracking per tenant, tenant-specific Knowledge Bases, memory isolation, and role-based model access are implemented through native AWS services. The model addresses key SaaS requirements: data protection between customers, QoS guarantees per tier, transparent cost attribution, and scalability without the operational complexity of dedicated systems.
AWS has provided sample code on GitHub demonstrating patterns for SaaS platforms, enterprise solutions with multiple business units, and managed services. The healthcare example with clinics and hospitals as tenants showcases the practical applicability of the architecture patterns to agentic AI.
Source: aws.amazon.com · Published June 23, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.1.