AWS shows how to build context-aware AI assistant on AgentCore
AWS published a technical guide on building a personal AI assistant that retains memory across conversations. The solution uses OpenClaw, an open-source agentic system, running on Amazon Bedrock AgentCore runtime. The example, named Sprout, is a gardening assistant that uses AgentCore memory to store user preferences and facts, allowing it to provide personalized advice without re-explaining…
Key points
- AWS demonstrates building a memory-enabled assistant using OpenClaw on Bedrock AgentCore runtime.
- The system uses Claude Haiku 4.5 for text and Claude Sonnet 4.5 for vision tasks.
- AgentCore memory stores user preferences and facts, enabling personalized responses across sessions.
The architecture is serverless and cost-effective, using consumption-based pricing that costs a few dollars a month for light use, compared to roughly $35/month for an always-on EC2 instance. It routes text queries to Claude Haiku 4.5 for speed and cost efficiency, while using Claude Sonnet 4.5 for vision tasks like diagnosing plants from photos. The system integrates with Telegram for user interaction and uses CloudFormation for deployment.
Key features include structured metadata for filtering memories and asynchronous extraction strategies that turn raw chat logs into durable, searchable records. This approach allows developers to create domain-specific assistants, such as fitness coaches or support bots, by swapping the persona and skills manifest while keeping the underlying infrastructure.
The story so far
10 episodes →- AWS shows how to build context-aware AI assistant on AgentCorethis story
Building a context-aware AI assistant on AgentCore and OpenClaw
AWS Machine Learning Blog · 6 October 2026
Loading the full article…
This text was published by AWS Machine Learning Blog and written by Thiago Verney. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Anthropic releases free crash course on building AI agents with Opus 5.5 · 1 src
- Meta's Muse AI permits explicit adult conversations, unlike other Meta apps · 1 src
- Elon Musk exposes Lego order details via Grok, users cancel purchase · 1 src
- OpenAI rolls out ChatGPT dots for Pro and Business Premium users · 2 src
- Former Plaud exec Peter Mo raises $26M for Ambi AI app that listens continuously · 1 src
Comments
via GitHub Discussions