Run an online store by texting commands to an agent
A reference build where an OpenClaw agent runs store API calls from the owner's Telegram texts, while a buyer-facing agent handles WhatsApp orders and status questions.
Done withOpenClaw
- What they did
- Claw Boutique is an AWS sample with two agents sharing one Store API. A Strands agent on Bedrock AgentCore, running Nova Lite, handles buyer WhatsApp chats. An OpenClaw agent on EKS sends the owner Telegram alerts for low stock and bad reviews, then runs the Store API calls for replies such as "restock hoodies 20" or "apologize". An apology sends a WhatsApp message and a refund email, marks the order refunded, and closes the escalation.
- How it went
- The post walks through the full order lifecycle working across the web store, WhatsApp, email, Telegram and an admin dashboard. It gives no performance or business figures, and it is a reference architecture, not a production deployment.
- Worth knowing
- One CDK deploy provisions everything, but a cold deploy takes about 25-30 minutes. It needs a WhatsApp Business account, a Telegram bot token and a verified SES address, and an EKS cluster and RDS database keep running until you destroy the stack.
Try it yourself with OpenClaw
Set yourself up as my store manager that I can text through [Telegram]. When I send a command like 'show today's orders' or 'mark order [number] as shipped', run the matching call against my [store platform] API and reply with the result. For anything that changes data, refunds money, or messages a customer, first send me a summary of exactly what you are about to do and wait for my yes. You're done when you can handle these three commands correctly: [command 1], [command 2], [command 3].
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: aws.amazon.com · Undated
Related use cases
Manage 18 agents from a Notion workspace
Notion used as the control plane for a fleet of 18 OpenClaw agents.
Run production deploys and port scans from Telegram with OpenClaw
A security professional uses an OpenClaw agent with read-only prod access, PR-only changes and manual approval for restarts to deploy changes and scan a VPS for exposed ports.
Run an autonomous blog content pipeline
An OpenClaw pipeline mines topics, researches, writes, validates, deploys and distributes content, with IndexNow submissions and analytics feedback; it reports 38 posts published.
Submitted by from search gap: seo
Run a solo company with eight agent departments
A founder set up CEO, CFO, COO, lawyer, accountant, marketing, CTO, and improver agents that share memory and consult each other.
Run bookkeeping and receipt matching in QuickBooks
A user wired AI into QuickBooks, SharePoint and email to attach receipts, forward invoices for checks, add vendors and request W-9s, replacing paid bookkeeping.
Custom agent
Click through apps and upload photos in the browser
Users report Claude in Chrome following a user's flow in an app, opening the file picker and uploading a photo, and once playing through online course tutorials.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.