An always-on Web UX Director
Our new site at cloudwalkeronaws.com has four AWS Strands Agents working on it behind the scenes. They analyse the live site, watch competitors, audit SEO, and deliver a prioritised list of improvements to our inbox every Monday morning. No human in the weekly research loop — just a human in the decision loop.
We built it this way because it's the same pattern we deploy for clients: define the job, break it into agent roles, ground each agent in the right data and context, schedule the execution, and route the output to a decision-maker. We wanted to run that pattern on ourselves before recommending any more of it.
"The research, analysis, and synthesis are handled by agents that never sleep and never forget to run. The human stays in the decision loop."
The four agents
How it works
Each agent is built with the AWS Strands Agents SDK, running Claude Sonnet on Amazon Bedrock. They execute as AWS Lambda functions triggered by Amazon EventBridge Scheduler — daily for the analysts, weekly for the synthesiser. Outputs land in Amazon S3 and ship via Amazon SES.
The agents have web browsing for competitor analysis, structured access to the site's HTML in S3, and a Bedrock Knowledge Base holding our brand guidelines, positioning, and target audience profiles. That grounding is what makes the suggestions specific to Cloudwalker instead of generic UX advice that could apply to anyone.
About the site itself
The site the agents operate on was built with Claude Code — Anthropic's CLI that lets you pair-program with Claude across a full project. Plain HTML, CSS, and JavaScript. No framework, no bundler, no CMS.
Claude Code treats the project as working context. It reads your CLAUDE.md, follows your brand guidelines, and keeps conventions consistent across dozens of HTML files. For a plain static site with no build step, it was near-perfect. Service pages, solution pages, contact form, success stories, insights — all consistent, on-brand, mobile-responsive — shipped in a handful of sessions.
This is worth naming because the site isn't the point; it's the substrate the agents operate on. The speed at which it came together is what freed us to spend the real engineering effort on the agents running on top of it.
What this means for our clients
We didn't build this as a novelty. It's a clean demonstration of an autonomous agent workflow: regular, structured analysis tasks that deliver actionable outputs to a human decision-maker on a schedule.
The same pattern applies wherever the work is recurring, structured, and research-heavy — customer data analysis, compliance monitoring, competitive intelligence, operational reporting. We've built exactly this kind of system before, and we know how to do it well.
"We build with these tools ourselves, not just recommend them to clients. That is the only honest way to stay ahead of it."
A shift, not a hype cycle
Claude Code isn't a toy. AWS Strands Agents aren't a prototype. These are production tools changing the economics and speed of software delivery right now, today. We've been in this space long enough to know the difference between a shift and a hype cycle — and this is a shift.
If you want to talk about what this means for your data infrastructure, your analytics stack, or your own agent strategy — you know where to find us.