Exploring the Intersection of OpenClaw and Agentic Commerce w Pod/Video/ Real-Life Demos🦞
By popular demand, let's dive deep deep into the salty waters of OpenClaw - the powerful and outside-the-box open-source agent to learn more about it's Agentic Commerce capabilities!
Welcome to a very special episode of the Retailgentic Podcast! In November of 2025, Peter Steinberger released ClawdBot into the world and here we are three short months later. Since then Peter renamed the project (at Anthropic’s request) to OpenClaw and was hired by OpenAI for an undisclosed amount of $.
In the last thirty days, on a daily basis, Retailgentic fans like YOU 🫵 have been relentlessly asking: “Scot, umm, haven’t you heard of OpenClaw?!” Embrace the CLAW🦞 Scot, come on!
Excuse Me, Do YOU Know About OpenClaw and Agentic Commerce?
Befuddled by the onslaught of inbound inquiries by aggressive Claw-sters🦞, I was on a trip to Silicon Valley last week, and stumbled upon what I thought must be the home of OpenClaw. Here I would find answers!
Alas, it turns out Luke had no idea what I was talking about, but did say I was the fifth person that week to ask about it and he didn’t know what was going on, but would I like to see the menu? 🤷♂️
Ryan Eade to the Rescue🦞
Rumor had it that my local tech-friend and super AI enthusiast Ryan Eade was deep into the OpenClaw world, so I scheduled a coffee meeting to see what that was all about and when he showed up singing this song dressed like this, I knew we had found our man!
The Mt. Everest 🌋 of Agentic Commerce Tests for OpenClaw…
After doing my research on X and the OpenClaw forums, I challenged Ryan and his armada of Claws to demonstrate these 3 super-hard Agentic Commerce tasks:
Task 1: Locate a the best price on a product both online and offline. **But an Important Twist: Have OpenClaw find detailed local inventory!
Local inventory at stores like Walmart, Target, BestBuy, Dick’s Sporting Goods are notoriously hard to execute on because an agent has to meticulously, find the local stores within an X mile radius, iterate through them, changing the store locator, executing the search, finding the item and pulling the availability and price. While this is a very parallelizable problem (agent swarm!), I have yet to see a demo of anything even close in browser-scraping agentic commerce (which OpenClaw is).
Task 2: Buy Something From Amazon 😱
As covered in deep detail in these virtual pages, Amazon is very very very very anti-bot. Could Ryan and his Claw defeat the tech magic that Amazon employs to ward off the evil Lobsterbots?
Task 3: BOSS LEVEL: VisionClaw!
I thought, hmmm, what could be harder than shopping on Amazon?! Then after hours of searching X, I found a hacker that had taken a pair of Ray-Ban META AI glasses, hacked his OpenClaw in as a personal voice assistant and then on top of that he created a shopping skill that lets him look at any physical object, perform an image/object search, speak back the results for further refinement, or you can order from that list verbaly.
Claw-powered Augmented Reality Real-time Agentic Commerce!?!
OH MY CLAWD🦞 !!
This would be awesome, but, I was very suspicious that the one demo video I found could be fake AI generated video - it’s hard to tell these days. I’ve seen these types of AR/VR demos and they are super slow, flakey and never as crisp as this demo. It felt too good to be true. Plus META glasses are notoriously closed, how is this possible?
I presented this to Ryan, he asked for a minute and feverishly texted his OpenClaw-powered agentic assistant and they return with a hardy: “Challenge: Accepted!”.
He didn’t know yet, but to make it EVEN harder, I threw a curveball in. Ryan’s Clawd would have to find my go-to hard to find-and-match product - Veggiedent dog treats. Those of you that have followed my first live streams of Agentic Browsers and other crawler-based systems, have seen them break instantly at the complexity of the seemingly simple Veggiedent dog treats. (It’s because they come in variations of dog size AND they have product variations fresh/something/something-else).
Note: They look harmless, but these VEGGIEDENT dog treats strike fear into the hearts of Agentic Commerce bots around the World!
OpenClaw Disclaimer and Skills Used In This Demo
In this podcast, Ryan takes us through all of the above. Can he pull it off? This is where I tell you, you have to watch or listen to find out!
Before we jump into how to watch/listen to this bonkers demo, first a couple of critically important notes:
Disclaimer: OpenClaw and Agentic Commerce Are Inherently Insecure.
Ryan is a security expert and while he implements some clever systems here to minimize the footprint, he did this in a very controlled environment, that wasn’t purely off-the-shelf OpenClaw, but some extra security rails were added. If you decide to experiment with OpenClaw and give it access to any credit card or bank details, you have been warned, this is highly risky.
If you want to learn more about why this is risky and why our friends at Nekuda don’t think it’s the future of Agentic Commerce, I highly recommend their substack on the topic here→
Some OpenClaw Skills Required
For those of you aware of the risks and looking to learn more at your own peril, here are the OpenClaw skills mentioned in this podcast. Disclaimer2 - we have not validated that these are risk free, a common attack vector on OpenClaw is prompt injection via skills, be careful, you are on your own, this is in no ways a valiation of these skills to be risk free, on top of an already risky system.
With that disclaimer, here are the skills:
Skill 1: Rye Universal Checkout
Details here. Rye offered the best solution for solving the Amazon problem.
Skill 2: SERP AI
Details here. This skill dramatically speeds up the execution of the OpenClaw agent in the Research→Find phase of the purchase cycle.
Skill 3: Buy-anything
Details here. After saying/or typing his bill to/ship to 1000 times, Ryan decided to find or build a skill that would persist that and make it easy. buy-anything does that.
Skill 4: VisionClaw
Finally, the skill that makes VisionClaw possible is here. Note: Requires Ray-Ban META glasses.
Where to Watch
Where to Listen
Addendum
Unfortunately due to our recording software, Streamyard, we could not get the Visionclaw demo to stream correctly, but I can 100% verify it worked flawlessly on the veggiedents the first time it had ever seen them. I literally had to pick my jaw up off the floor, there were 2-3 times in this whole experience where I felt like I was on the edge of the reality I knew and a new one that - it’s hard to explain how mind-blowing some of these experiences are.
I asked Ryan to record a stand-along video so you could see the entire experience and here it is:
And here’s the output on his Claw’s (RonClawd) slack channel:
