Inspiration
Business buying is more than adding products to a cart. Teams must balance budgets, compatibility, stock, delivery dates, discounts, and payment terms. I built Northline to show how an AI agent can manage that complexity while keeping people in control.
What it does
Northline is a fictional B2B office supplier powered by WebMCP. An agent can turn a request like “equip a room for six people” into a complete procurement plan, compare alternatives, check constraints, update the visible cart, and request approval before placing the order.
How I built it
I created a browser-based storefront with 63 products, live inventory, quantity pricing, account discounts, delivery options, and B2B checkout. Sixteen WebMCP tools connect the agent directly to the catalogue, cart, account, and ordering workflows.
Challenges I ran into
The hardest part was making the demo realistic without making it overwhelming. I also had to design tools that produce clear errors, recover gracefully, and separate reversible actions from binding purchases.
Accomplishments that I'm proud of
I built a complete human and agent buying flow with meaningful trade-offs, not just product search. The agent handles the detailed work, while the buyer sets priorities and explicitly approves the final order.
What I learned
Structured browser tools make agents far more reliable than screen scraping or brittle click automation. I also learned that the best agent does not hide uncertainty. It explains constraints and asks people to resolve important trade-offs.
What's next for B2B Procurement
Next, I would add approval chains, supplier integrations, negotiated catalogues, multi-warehouse inventory, audit logs, and ERP connectivity. These additions would bring the same collaborative experience to real procurement teams.
Built With
- codex
- css
- html
- javascript

Log in or sign up for Devpost to join the conversation.