PulseFlow Tecnologia
← Insights
·5 min read·product-update·aws·bedrock-agentcore·rag
PulseFlow Tecnologia

What AWS Bedrock AgentCore changes for enterprise agents

AWS shipped, in quick succession, two features that together redesign what it means to run an AI agent in production on Bedrock: Web Search on AgentCore and Managed Knowledge Base. Separately, each solves a specific problem. Together, they form AWS's answer to a particular question: how does an agent know what it needs to know, without the engineering team building that from scratch.

Web Search: grounding in the real world

Amazon Bedrock AgentCore is AWS's managed platform for running AI agents with tool access inside the customer's AWS environment. The Web Search feature added to that platform solves a specific problem: an agent only knows what was in the model's training data - and that ages fast. The tool returns "snippets, URLs, titles, and publication dates" for the agent to reason over, built on the same search infrastructure that already powers Alexa+, Amazon Quick, and Kiro. Importantly, it keeps "zero data egress" from the customer's AWS environment, and uses the Model Context Protocol to communicate - available in US East (N. Virginia), at $7 per 1,000 queries.

Managed Knowledge Base: RAG without reinventing the pipeline

The second announcement attacks a problem that precedes external grounding: how to connect an agent to a company's own internal data without every team building its own RAG pipeline from scratch. Managed Knowledge Base ships with six prebuilt native connectors (Amazon S3, SharePoint, Confluence, Web Crawler, Google Drive, OneDrive), a "Smart Parsing" feature that automatically picks the chunking strategy per content type, and an "Agentic Retriever" that performs multihop retrieval - decomposing a complex question into multiple queries against multiple knowledge bases. The service is available in eight regions (including GovCloud), with no-upfront-commitment pricing based on indexed data stored and number of retrievals, and integrates with LangChain, CrewAI, and LlamaIndex via MCP.

What actually changes

Before these two launches, building a production agent on AWS meant solving, on your own, both grounding in current knowledge and the RAG pipeline over corporate data - each with its own engineering and maintenance cost. With Web Search and Managed Knowledge Base, AWS packages both as a managed service, priced by usage, with native MCP integration.

For anyone evaluating building agents on Bedrock, the takeaway is direct: the complexity that used to sit with the engineering team (connectors, chunking, embeddings, re-ranking, keeping external knowledge current) now belongs to the platform itself. That lowers the barrier to putting an agent into production - but it also means the company's data architecture choices become more tightly coupled to AWS's specific choices.

Sources