Articles
Amazon Bedrock AgentCore: the runtime foundation for building, deploying and scaling AI agents
The Next Chapter of Enterprise AI: From Models to Agents
Large language models (LLMs) were only the opening act. The true transformation begins with autonomous agents: intelligent systems that can reason, take action, and connect directly to business systems.
To power this shift, AWS has introduced Amazon Bedrock AgentCore, a managed service that provides the runtime foundation for building, deploying and scaling AI agents. For enterprises ready to step forward from experimentation to execution, AgentCore offers a secure, flexible, and scalable path forward.
As Joris Conijn, CTO of AWS at Xebia, puts it “AgentCore is to AI agents what the AWS Landing Zone was to the cloud. It gives organizations a predictable, scalable foundation to innovate with confidence, without worrying about the infrastructure.”
Agentic AI with Amazon Bedrock AgentCore
See how Amazon AgentCore brings Agentic AI to life. On-demand webinar featuring Brajendra Singh, GenAI Lead at AWS, and Joris Conijn, CTO AWS at Xebia.
Watch now
On-Demand Webinar
Why AgentCore is a Game-changer
Unlike configuration-based Bedrock Agents, to Bedrock models, AgentCore is built on openness and interoperability. It allows developers to build once and deploy anywhere, across popular open-source frameworks such as LangChain, LangGraph, or Strands, and with any model provider from Anthropic to Cohere or Mistral.
According to Brajendra Singh, Principal Solution Architect and AWS GenAI Go-to-Market Lead, this open design is intentional:
“We wanted to make it frictionless for developers to move from idea to implementation. AgentCore lets you bring your preferred frameworks and models while AWS takes care of the scale, the security, and the lifecycle management.”
Brajendra Singh
GenAI Lead at AWS
The Six Components of AgentCore
AgentCore brings together six managed components that handle the entire agent lifecycle, from orchestration to governance.
- Runtime: Serverless, Scalable, Secure.
Powered by a serverless microVM architecture based on Firecracker, it ensures every user session runs in isolation. It guarantees enterprise-grade security and predictable scalability without the need to manage servers. - Gateway: the Bridge to Business Systems.
Agents act through tools, and the Gateway is the secure bridge that connects them to APIs, databases, and SaaS platforms securely and uniformly. With MCP-compatible interfaces and semantic search capabilities, enterprises can build an internal tool hub where agents discover, reason, and invoke the right tools for any task. - Identity: Trust Built In
This component manages both inbound and outbound authentication, allowing secure agent-to-agent communication and API access. By centralizing credentials and abstracting them from the code,it enforces zero-trust principles, while meeting enterprise security standards. - Memory: from Conversation to long-term Knowledge
AgentCore’s built-in memory services store and retrieve both short-term conversational context and long-term semantic knowledge. This enables agents to recall user preferences, learn over time, and provide context-aware personalization, all without developers manually managing vector databases. - Observability: See How Your Agents Think
With native observability, every interaction, decision, and tool invocation is logged in Amazon CloudWatch, and can be integrated into any OpenTelemetry-compatible observability platform. This gives developers visibility into how agents reason, making debugging and compliance straightforward. - Built-in Tools: Ready to Use
AgentCore enables you to hit the ground running with out-of-the-box tools like Code Interpreter and Browser, extending the agent’s reach to data analysis, automation, and web interaction. Developers can also publish custom tools via the Gateway for specific enterprise needs.
AgentCore in action : From Reactive to Proactive AI
During the Xebia–AWS webinar, live demos showcased how AgentCore turns complex orchestration into deployable intelligence:
- The Talent Finder Agent: A multi-agent setup where an Orchestrator Agent coordinates with specialized Match and Availability Agents via the Gateway. It autonomously searches for candidates, checks their availability, and schedules meetings, all through seamless inter-agent communication.
- The SRE Incident Response Agent: When a CloudWatch alarm triggers, the agent automatically investigates logs, identifies root causes, and suggests fixes, slashing the mean time to recovery (MTTR).
“This is where agents move from being assistants to active participants.AgentCore enables systems that can take responsibility, not just generate answers.”
Joris Conijn
CTO AWS at Xebia
Governance, Guardrails, and Responsible AI
Innovcation needs to be balanced with control. AgentCore supports Bedrock Guardrails out of the box, allowing teams to filter unsafe or biased outputs and enforce data access policies.
The Identity and Gateway layers also manage data provenance, authorization, and content grounding ensuring every agent’s response is both fact-based and compliant.
“Governance in AI isn’t optional,” noted Singh. “AgentCore embeds security, observability, and responsible AI controls into the agent lifecycle, so enterprises can focus on innovation without losing oversight.”
The Road Ahead: Agentic AI as the New Normal
AgentCore marks a fundamental evolution in enterprise-AI: from prompting to orchestrating, from models to ecosystems.
Just as AWS Landing Zones standardized cloud foundations, AgentCore will define how organizations deploy secure, scalable agent infrastructures. It’s the missing layer that makes agentic AI production-ready.
“With platforms like AgentCore, we’re moving from experimentation to execution. This is the foundation for the next decade of enterprise AI.”
Joris Conijn
CTO AWS at Xebia
Ready to start with AgentCore on AWS?
After the preview phase, AgentCore has been made available to the general public in October 2025. Early experimentation offers organizations a first-mover advantage in understanding how to operationalize agentic AI at scale. Explore the official documentation of AWS and sample repositories to start building today and see how Xebia can accelerate your Agentic AI adoption on AWS.
Accelerate your Journey with Xebia
Partnering with Xebia allows you work with one the first AWS Premier Partners, and one of the first to achieve the AWS Digital Sovereignty Competency, Xebia helps enterprises design, secure, and scale AI solutions on AWS from foundational models to production-grade agentic systems. Partner with Xebia to get started!
Our Ideas
Explore More Articles
Multi-Agent Systems and Gemini Enterprise: Intelligence as Code
Agentic systems combine reasoning with action, marking the next step in AI’s evolution. The shift from monolithic agents to orchestrated multi-agent...
Timothy van der Werf
Google Cloud: AI as a Change Management Imperative
AI adoption is as much about people and processes as it is about technology. By combining Google’s advanced AI platforms with a structured change...
Walter van der Scheer
The Economic and Strategic Impact of Agentic AI: Beyond Efficiency to Structural...
Walter van der Scheer
Contact