AWS guide: building ambient agents on Bedrock AgentCore with human-in-the-loop
AWS ML Blog · rss · 2026-10-02
Core idea
An AWS ML Blog post explains how to build ambient agents on Amazon Bedrock AgentCore — agents that listen to event streams (file uploads, schedules, DB changes, alerts) and act automatically, instead of waiting for a user-initiated chat prompt, and can handle many events in parallel.
Key design
- Ambient signals map event sources to agents. With autoExecute: false, jobs land on a Jobs page for human review first (safe default); with autoExecute: true, the agent runs immediately and only interrupts a human by calling askhuman.
- A single askhuman tool: the agent pauses for clarification or approval and resumes once answered — lowering production risk and building trust.
- Architecture: S3 notifications / EventBridge / Lambda for events, AgentCore Runtime for container-based long-running agent hosting with session isolation (15-min cap per turn), DynamoDB for state — fully serverless. Webhooks and DynamoDB streams are provided as extension points.
- The reference implementation uses Claude Sonnet 4.5 on Bedrock; switching models is a one-line config change.
More from coding & agent
- Dev Has Claude Build a Custom Jump Animation Editor to Speed Up Game Feel Iteration — andrew_n_carr · 2026-10-02
- Translating an entire book with DeepSeek: pennies and under an hour, decent quality — teortaxesTex · 2026-10-02
- OmniSeek turns Omni-LLMs into agents that actively seek audio-visual evidence — Haibo Wang · 2026-10-02
- Microsoft's ActiveSaddler Uses Automated Curriculum Learning to Boost Agent Harnesses by 7.5 Points — microsoft · 2026-10-02
- Alibaba's PoS Maintains Explicit Belief States to Fix Long-Horizon Agent 'Belief Trapping' — alibabagroup · 2026-10-02
- IntentFlux Benchmarks 'Intent Drift' in LLM Agents: Scores Fall from 0.476 to 0.384 as Users Change Their Minds — Yanjie Zhang · 2026-10-02