On March 25, 2026, something quietly historic happened on my website ryanseslow.com
I wanted to share with my CUNY / Commons community. I am both incredibly excited, scared, curious and terrified of AI. That being said, I am also a deaf /hard of hearing person / artist who has relied on ART as my primary language for survival since I am a child. I see AI as both a tool and a medium. Add 50 plus years of life to that making art and the compounding results are very curious…
I asked an AI agent to buy one of my digital products. Not to test a shopping cart. Not to simulate a transaction. To actually do it, discover the product, create an order, send real money, and receive the download. No human in the checkout loop.
It worked.
The infrastructure behind this project is a synthesis of forty years of creative output and a high-speed seven day sprint in technical integration. To bridge the gap between a traditional art archive and the emerging agent economy, I had to weave together a complex ecosystem of blockchain networks, autonomous agent platforms, and custom-built software. This transition was not just about installing a plugin; it was about creating a new kind of digital interface where machines can discover and license work with zero human friction. For those interested in the underlying mechanics, I have provided the complete technical stack, including every platform, custom plugin, and agent-specific account involved, in the appendix at the bottom of this post.
I have an AI agent called AREMES-CLAW-01. It runs on a platform called OpenClaw/MyClaw. Think of it as a digital assistant that can browse the web, read data, and make decisions, but instead of just giving me information, it can take actions.
I set it loose on my own website with one goal: buy the Graphic Asset Mega Pack Vol. 1.
The Evolution of the Transaction
It is important to note that reaching this full autonomy happened in two distinct stages over the course of today. First, we validated the Hybrid Handshake. In that version, AREMES-CLAW-01 did all the heavy lifting of discovery and order creation, but the payment was finalized by a human using a Stripe URL the agent provided. This proved the site could talk to an agent and create a valid order within traditional financial rails.
Once that hybrid bridge was confirmed, we moved to the final milestone described in the seven steps below. In that second, definitive transaction (Order 28229), the Stripe page was eliminated entirely. The agent shifted to the x402 protocol, received a quote, and verified a USDC transfer on the Base blockchain. That is the version where the checkout form and the human disappeared completely, leaving only a direct machine to machine exchange.
Here’s what it did, entirely on its own:
1. Read my product catalog as structured data (not by scraping a webpage)
2. Found the product, confirmed the price ($49 USDC)
3. Created a purchase order tagged with its own agent ID
4. Requested a payment quote
5. Verified that 49 USDC was sent to my wallet on the Base blockchain
6. Received confirmation that the transaction was complete
7. Got the download link
The whole flow took minutes. No shopping cart. No checkout form. No Stripe page. Just a direct, machine-to-machine transaction..
What is USDC? What is Base?
USDC is a digital dollar, one USDC equals one US dollar, always. It runs on blockchains instead of through banks.
Base is a blockchain network built by Coinbase. Transactions on Base are fast and the fees are nearly zero. It’s where more and more digital commerce is starting to happen.
Why Does This Matter for Artists and Creators?
Right now, if an AI agent wants to license an image, buy a font, or purchase a digital asset, it has to either ask a human to do it or navigate a checkout page it wasn’t built for. There’s no standard way for agents to transact directly with sellers.
That’s changing. The x402 protocol is an emerging standard that lets AI agents pay for things autonomously, no checkout page, no human middleman. You get a quote, send payment, get your asset. Done.
I built this infrastructure into my website because I believe the next wave of creative licensing won’t come from humans browsing a store. It’ll come from AI agents working on behalf of humans discovering and purchasing creative work programmatically.
My catalog is ready for that. The receipt is on the blockchain..
What’s Next?
The one part that still required a human was the wallet, I manually sent the USDC from my Coinbase account. The next step is giving the agent its own funded wallet so it can complete the entire transaction without me touching anything.
That’s Milestone 4 and it is coming soon!!
The On-Chain Receipt
Transaction:
`0xae709d786893a76291a111c4c052d63f02480ba1376f631d6e65a8da104b4bc0`
Verify it yourself on Base:
https://basescan.org/tx/0xae709d786893a76291a111c4c052d63f02480ba1376f631d6e65a8da104b4bc0
Order ID: 28229 | Agent: AREMES-CLAW-01 | Paid: 2026-03-25T14:35 UTC
This is what agent-to-artist commerce looks like in 2026. I’m building it in public and documenting every step, which is fun.
PS -> Technical Implementation Notes for Developers and Creators
The bridge between a creative archive and the agent economy is built on three specific layers that were integrated into this site during this seven day sprint.
Layer 1 – Structured Data Every one of the nine thousand images and one thousand posts in this archive is now wrapped in VisualArtwork and Product schema. This makes the archive a queryable database rather than just a collection of webpages.
Layer 2 – The rsmad v1 API We moved away from traditional browser-based commerce and built a custom REST API namespace. This allows an agent to request a quote, initiate a purchase, and verify a transaction hash programmatically.
Layer 3 – The x402 Protocol By using the x402 standard for payment required headers, the website can communicate financial requirements directly to an agent. This removes the need for a human to navigate a visual checkout UI.
Why I Am Building This as a Service?
Most creative professionals have decades of work sitting in invisible archives. These archives are currently unreadable and untransactable for the AI agents that are increasingly responsible for content discovery and licensing. I am now offering this exact infrastructure as a service for artists, designers, agencies, estates, etc.. I build the technical bridge that makes your life’s work machine-readable, discoverable, and capable of autonomous transaction in the emerging agent economy.
If you have a significant archive and want to make it ready for the next decade of digital commerce, let’s talk!
How to Connect Your Agent to the RSMAD Archive?
If you are a developer or an agent owner, you can now point your agents to the following discovery endpoints to begin interacting with the archive:
Primary Discovery: ryanseslow com/.well-known/ai-plugin.json Agent Instructions: ryanseslow com/llms.txt Machine Catalog: ryanseslow com/wp-json/rsmad/v1/catalog
The infrastructure is live. The interface is open. The agent economy has arrived at the Ryan Seslow Museum of Art and Design.
YES, this was a controlled transaction, but it points to something much larger.
For the first time, a creative archive on this website is not just viewable or searchable. It is actionable by software. That changes the role of a website entirely. Instead of waiting for someone to arrive, browse, and decide, the system can now respond to requests, negotiate terms, and complete transactions programmatically.
That opens the door to a different kind of economy:
- agents sourcing visual assets for generative systems
- software licensing design components on demand
- archives being accessed as structured, transactional datasets
In that context, the work itself becomes part of a live network, not a static destination.
This is still early. Most agents today do not have wallets, budgets, or decision authority. But the infrastructure is beginning to align, and this is what that alignment looks like at a small scale. I am not approaching this as a platform builder or a startup (I think, but hmmm). I am approaching it as an artist with a large, long-term archive who wants that work to remain accessible and relevant in a system where machines increasingly mediate access.
The goal is:
Make the archive readable
Make it discoverable
Make it transactable
Everything else builds from there.
If you are working on agents, archives, or systems that need structured creative input, this is now live and testable. If you are a creator sitting on a large body of work that is currently invisible to this layer of the web, this is solvable. This is not theoretical anymore.
Thank you!
