New in Amazon Bedrock AgentCore: Build agents with broader knowledge and continuous learning

Building intelligent agents requires access to comprehensive, accurate information. The expanded knowledge connectivity in AgentCore addresses a critical need for enterprises that want their agents to operate with both internal expertise and external context. By combining organizational data with web and paid knowledge sources, agents can make more informed decisions and provide more valuable responses to users. This multi-source approach reduces hallucinations and improves reliability in production environments where accuracy matters.

Deploying agents to production introduces complexity that's difficult to manage without proper tooling. AgentCore's improved troubleshooting capabilities give teams the visibility and diagnostics they need to maintain agent reliability at scale. When agents encounter errors or behave unexpectedly, teams can now quickly understand the underlying cause. This might be a failing API connection, a knowledge gap that prevented the agent from answering a question, or a reasoning mistake in the agent's decision-making process. With these insights, teams can prioritize fixes and iterate rapidly to improve agent performance.

One of the biggest challenges with agent deployment is maintaining appropriate oversight as agents handle increasingly important business decisions. AgentCore addresses this with governance controls that scale alongside agent capabilities. Rather than managing agents as one-off deployments with ad-hoc restrictions, the platform provides a unified framework that grows with your agent ecosystem. This means organizations can confidently expand agent responsibilities while maintaining compliance, security, and alignment with corporate policies. The controls work across multiple agents and can be updated centrally, reducing administrative overhead and ensuring consistent governance across all agent deployments.

Speed matters in AI development. Organizations that can build and deploy agents faster gain competitive advantages by automating business processes earlier and more effectively. AgentCore's new capabilities remove traditional bottlenecks in agent development. Instead of waiting for security teams to approve each new agent, governance controls are built in and can be verified automatically. Instead of struggling to understand what went wrong in production, diagnostic tools provide immediate clarity. These changes compound to create significantly faster development cycles where teams spend less time troubleshooting and more time innovating.

Amazon Bedrock provides a managed platform for building with foundation models, and AgentCore extends this with agent-specific capabilities. The new features work seamlessly within the Bedrock ecosystem, allowing teams to combine improved agent development with Bedrock's model management, fine-tuning, and customization tools. This integration means organizations can use the same platform for both foundational model operations and agent orchestration, simplifying architecture and reducing the number of tools teams need to manage.

For enterprises, deploying agents at scale requires balancing capability with control. A financial services company might want agents that can answer customer questions about accounts and products, but needs absolute confidence that agents won't provide regulatory-violating advice. A manufacturing organization might deploy agents to optimize supply chains, but needs to understand every decision the agent makes. These new AgentCore capabilities support both scenarios. By providing better knowledge integration, production visibility, and scalable governance, the platform enables enterprises to realize agent benefits while maintaining the oversight and compliance that larger organizations require.

Agent development is still a relatively new frontier for most organizations, and best practices are still emerging. AWS's investment in troubleshooting, governance, and knowledge integration suggests a pragmatic approach based on real-world customer feedback. These aren't theoretical improvements but responses to actual challenges that enterprises face when building and operating agents. As the market matures and organizations deploy more agents, expectations for reliability, auditability, and governance will only increase. AgentCore's new capabilities are designed to meet those expectations from day one.