HEMA Built an Enterprise AI Assistant with MCP and Amazon Bedrock AgentCore
AWS ML Blog · rss · 2026-09-24
AWS ML Blog details how Dutch retailer HEMA (750+ stores) built HAL, an internal AI assistant on Amazon Bedrock AgentCore and MCP, consolidating knowledge scattered across wikis and portals. Key architecture: local Strands tools query Bedrock Knowledge Bases directly; an MCP connection to AgentCore Gateway (auto-generated from OpenAPI specs and Lambda) serves live APIs and service-catalog lookups — no custom MCP server code. Security uses Microsoft Entra ID OAuth, JWT inbound auth, an outbound OAuth2 token vault, and AD-group access control (read-only today). AgentCore Runtime, Memory, and Guardrails provide hosting, memory, and filtering with EU regions and Dutch support. Answers that once took an afternoon of portal-hopping now take seconds inside IDEs and chat.
More from coding & agent
- QueueSim MCP Connector Lets AI Agents Run M/M/c Queue Simulations Across Four Scenarios — modelcontextprotocol · 2026-09-24
- predictfun-mcp Gives AI Agents Structured Access to $1.5B+ Prediction Market on BNB Chain — modelcontextprotocol · 2026-09-24
- Sandboxing AI agents with FnCall: network-only syscall limits and the MIG gotcha — sloppenheimer · 2026-09-24
- Merge launches Universal Context Layer, a shared company brain for every AI agent — shensi · 2026-09-24
- Lydia Hallie: Claude Design produced all video visuals in ~1.5 hours — lydiahallie · 2026-09-24
- Cline Desktop v0.0.34 adds worktrees, GitHub PR status, and parallel subagents — cpaik · 2026-09-24