01
Agentic Commerce
Agents that find products, build carts and reorder through commerce APIs, with permissions scoped to the task.
JohnathonMathews
-
OpenAI -
Anthropic -
Dify -
Langflow -
n8n -
Windmill
Fig. 01 · Agents on real orders, live
Hi, I’m Johnathon.
AI and agentic commerce · Phoenix, AZ
I put AI to work on real orders, prices and inventory. Guardrails built in from day one.
About
Learning AI in the Work,
Not Around It.
I’m a software architect in Phoenix working where commerce platforms meet AI: search, personalization and agents that act on real orders and inventory. Fifteen years of commerce taught me that data quality and permissions decide whether an AI feature can be trusted.
I’m learning by bringing AI into my day-to-day work, from development and research to building my own products. Some things work well. Others take more work than the demo suggests.
Today, I lead the technical strategy for my teams, connecting what we build to what the business needs. As AI changes what’s possible, I keep revisiting our assumptions, testing new approaches, and helping the team decide what’s worth pursuing.
Areas of focus
Where I’m sandboxing right now. Each area is built and tested and proven. Constantly learning and growing.
02
MCP Servers
Exposing catalog, order and inventory data to agents through the Model Context Protocol, one tool at a time.
03
Agent to Agent
Agents handing work to other agents: who owns the task, what gets passed along and where a person signs off.
04
Chunking Processes
Breaking long commerce workflows into small, checkable steps an agent can run, verify and retry safely.
05
Edge Personalization
Per-shopper content decided at the edge, so pages stay fast and caches stay honest.
06
Guardrails and Evaluations
Policy gates, approvals and evals so a bad answer never becomes a bad order.
Working set
Leverage
ChatGPT
Codex
Claude Code
Hermes Agent
Superdesign
ElevenLabs
Dify
MCP
Notes
Writing on AI
Field notes in progress, forever behind but forever learning.
New articles are on the way. Check back soon.
Schedule time
Book a Call
30 minutes. Real conversation. Useful next steps.
Checking availability...