Amazon Bedrock AgentCore now available in four additional AWS Regions
Run production AI agents with lower latency in Southeast Asia and Southern Europe—full AgentCore capabilities available from day one in all four new regions.
View original announcement →Visual Summary
graph TD
A{{Amazon Bedrock AgentCore}}:::announced
B([Agent Runtime]):::feature
C([Identity & Access]):::feature
D([Gateway & Tools]):::feature
E([Memory & Sessions]):::feature
F([Policy & Security]):::feature
G([Observability]):::feature
H((Developers)):::external
I((New Regions)):::external
H ==>|"deploy agents"| A
A -->|"hosts"| B
A -->|"authenticates"| C
A -->|"connects tools"| D
A -->|"persists context"| E
A -->|"enforces policies"| F
A -.->|"traces & debugs"| G
I ==>|"lower latency"| A
classDef announced fill:#ff9900,stroke:#ec7211,color:#fff,font-weight:bold
classDef compute fill:#e3f2fd,stroke:#1565c0,color:#1565c0
classDef storage fill:#e8f5e9,stroke:#2e7d32,color:#2e7d32
classDef feature fill:#fff3e0,stroke:#e65100,color:#e65100
classDef external fill:#f5f5f5,stroke:#616161,color:#616161
What's New
Amazon Bedrock AgentCore, the platform for building, connecting, and optimizing production AI agents, is now available in four additional AWS Regions: Asia Pacific (Bangkok), Asia Pacific (Malaysia), Europe (Milan), and Europe (Spain). This expansion brings the full suite of AgentCore capabilities—including agent runtime, identity and access control, policy management, session persistence, tool connectivity, and observability—to customers in these regions at launch. Organizations in these geographies can now deploy agents closer to their end users, reducing latency and meeting local data residency requirements.
How It Works
- AgentCore Runtime provides a secure, serverless hosting environment for agents and tools, supporting both direct code deployment for rapid iteration and container-based deployment for maximum control, billing only for active CPU and memory consumption rather than pre-allocated resources.
- AgentCore Identity manages inbound and outbound authentication, enabling agents to securely access third-party and AWS services without developers having to hand-roll credential management or access control logic.
- AgentCore Gateway acts as a managed connectivity layer that exposes APIs, Lambda functions, and MCP servers to agents at scale, handling authentication and authorization on every tool call.
- AgentCore Memory provides both short-term session persistence and long-term memory resources, enabling context-aware interactions across agent turns and sessions.
- AgentCore Policy enforces access policies verified by automated reasoning (the same technology underlying IAM and S3), ensuring security controls are applied at the infrastructure layer that agents themselves cannot bypass.
- AgentCore Observability captures step-by-step traces of agent execution—what was called, in what order, and where failures occurred—enabling faster debugging and continuous optimization.
- AgentCore Harness provides a single API call interface to define and invoke agents with a managed loop that includes built-in memory, tools, and isolated execution environments, and carries no additional charge beyond underlying resource costs.
- The platform is framework-agnostic, supporting LangChain, OpenAI Agents SDK, Claude Agent SDK, Strands SDK, and custom frameworks, with the same code deployable from local development to production.
Why It's Important
- Reduced latency for regional users: Customers in Southeast Asia (Thailand, Malaysia) and Southern Europe (Italy, Spain) can now run agents in-region, eliminating cross-region round trips that add meaningful latency to interactive agentic workflows.
- Data residency and compliance: Many enterprises in the EU and APAC operate under strict data sovereignty regulations; running AgentCore workloads within local AWS Regions helps satisfy these requirements without architectural compromises.
- Accelerated enterprise adoption: The availability of the full capability set at launch—rather than a partial rollout—means teams in these regions can immediately build production-grade agents without waiting for feature parity.
- Infrastructure-layer security: Because security policies are enforced at the platform level using automated reasoning, organizations gain a verifiable security posture that doesn't depend on agent code behaving correctly.
- Operational efficiency: Industry data cited by AWS suggests AgentCore reduces time spent on infrastructure and integrations significantly compared to custom development, freeing engineering teams to focus on agent logic and differentiation.
How It's Different
- Security enforced below the agent layer: Unlike middleware or SDK-level guardrails that an agent could theoretically circumvent, AgentCore enforces access policies at the infrastructure layer using automated reasoning—the same mechanism behind IAM and S3—making security non-bypassable by agent code.
- Active-consumption billing model: Traditional compute charges for pre-allocated resources even during I/O wait; AgentCore Runtime bills only for actual CPU and memory consumed per second, which is particularly advantageous for agentic workloads that spend 30–70% of their time waiting on LLM responses or external API calls.
- Framework and model agnosticism: Most managed agent platforms are tightly coupled to a specific framework or model provider; AgentCore is explicitly designed to work with any framework and any model, preserving optionality as the ecosystem evolves.
- Full capability parity at regional launch: New regions receive the complete AgentCore feature set (runtime, memory, gateway, identity, policy, observability, evaluations, optimization) rather than a staged rollout, reducing friction for teams building in those geographies.
- Integrated tool connectivity with per-call auth: AgentCore Gateway handles authentication and authorization on every individual tool call (MCP servers, Lambda, internal APIs), a capability that typically requires significant custom plumbing when built independently.
When to Prefer It
- Regulated industries in the EU or APAC: Organizations in financial services, healthcare, or public sector in Italy, Spain, Thailand, or Malaysia that must keep data and compute within specific geographic boundaries should use the newly available regional endpoints.
- Latency-sensitive user-facing agents: Applications such as customer support bots, real-time coding assistants, or interactive data analysis tools where sub-second response times matter will benefit from running agents in the region closest to end users.
- Teams adopting multiple frameworks or models: Engineering organizations that want to experiment with different agent frameworks (LangChain, Strands, OpenAI Agents SDK) or swap foundation models without rewriting infrastructure should prefer AgentCore's framework-agnostic runtime.
- Enterprises needing auditable agent security: When compliance or audit requirements demand verifiable proof that agents cannot exceed their granted permissions, AgentCore's automated-reasoning-backed policy enforcement provides a stronger guarantee than application-level controls.
- Teams scaling from prototype to production: Organizations that have built proof-of-concept agents and need to add session persistence, observability, tool connectivity, and access control without rebuilding from scratch will find AgentCore's modular, additive architecture well-suited to this transition.
- Cost-sensitive agentic workloads: Workloads with high I/O wait ratios (LLM-heavy pipelines, multi-step tool-calling agents) benefit from AgentCore Runtime's active-consumption pricing model versus paying for idle compute on traditional instance-based infrastructure.
Availability
- Generally Available: Amazon Bedrock AgentCore is GA in the four newly added regions as of July 1, 2026, with the full feature set available at launch.
- New regions: Asia Pacific (Bangkok / ap-southeast-7), Asia Pacific (Malaysia / ap-southeast-5), Europe (Milan / eu-south-1), and Europe (Spain / eu-south-2).
- Existing regional coverage: AgentCore Runtime and several core features are available across 20 regions including US East (N. Virginia, Ohio), US West (Oregon), multiple European regions, APAC regions, Canada, South America (São Paulo), and AWS GovCloud (US-West).
- Feature availability varies by region: Some capabilities—AgentCore Payments (preview), AWS Agent Registry, AgentCore Optimization, and AgentCore Evaluations—are available in a subset of regions; consult the Supported AWS Regions documentation for the full matrix.
- Pricing model: Consumption-based with no upfront commitments or minimum fees; AgentCore Harness carries no additional charge (pay only for underlying resources); AgentCore Runtime bills per second based on actual CPU consumption and peak memory; other features are priced independently and can be used à la carte.
- Free Tier: New AWS customers receive up to $200 in Free Tier credits; AWS Agent Registry is free during its preview period.
- Resources: Product page at aws.amazon.com/bedrock/agentcore/, Developer Guide at docs.aws.amazon.com/bedrock-agentcore/, and region availability table at the AgentCore Developer Guide regions page.
