AffiliateTresur

AffiliateTresuroperates as an agent-native affiliate discovery platform, bridging human browsing and AI interactivity

Inspiration

Conventional affiliate discovery platforms are designed exclusively for human navigation through ad banners, cluttered lists, and search bars. However, as local client assistants and autonomous AI agents evolve into the primary portal for research and shopping, web content must become programmatically actionable for machines as well as visually readable for humans. Driven by this shift, we set out to create an agent-native affiliate marketplace. Our goal was to enable seamless visual browsing for users while empowering local AI agents to query structured product catalogs, discover tools, and deliver verified deals using modern browser standards.

What it does AffiliateTresur.

How we built it

Frontend: Built with vanilla HTML5, modern CSS, and JavaScript to keep asset delivery lightweight and responsive.
Agent Interface: Implemented declarative WebMCP tool handlers that bind directly to DOM forms and programmatic schemas, responding with structured JSON when invoked by browser-level AI agents.
Edge Hosting & Delivery: Configured for edge deployment via Cloudflare using Wrangler, delivering static assets globally with instant edge caching.
Version Control & CI/CD: Maintained via GitHub with automated Cloudflare Worker build triggers and custom domain routing on affiliatetresur.com.

Challenges we ran into

Edge Environment vs. Browser Code: Reconciling Cloudflare Worker deployment pipelines with a purely client-side static application required fine-tuning wrangler.jsonc asset routing rather than compiling edge server handlers.
Build Configuration Hurdles: Diagnosing configuration file syntax issues—such as multi-root JSON parsing bugs in package.json—and configuring compatibility dates under automated Wrangler pipelines.
Standardizing Agent Discovery: Aligning custom mock product query filters (spanning hardware, mobile tech, and automotive accessories) with strict WebMCP schema definitions so simulated agents could query data cleanly without format errors.

Accomplishments that we're proud of

Seamless Tool Discovery: Successfully implemented in-browser agent discovery, enabling Chrome DevTools agent simulators to detect and execute catalog queries instantly.
Zero-Bloat Architecture: Delivered a full-featured, functional prototype without heavy framework overhead, ensuring sub-second load times and zero unnecessary dependencies.
Live Domain Deployment: Successfully deployed and routed the platform to the live production custom domain affiliatetresur.com.

What we learned

Agent-First Web Design: Structuring data and forms for both browser DOM events and agent-invoked events requires thinking about web applications as APIs and visual dashboards simultaneously.
Declarative Browser Standards: Practical hands-on experience navigating emerging browser AI APIs and testing agent interaction pipelines locally and in staging environments.
Cloudflare Static Assets on Workers: Mastering modern Wrangler asset-binding patterns to serve high-performance client code reliably at the edge.

What's next for AffiliateTresur

Live Affiliate API Feeds: Connect the platform beyond mock datasets to real-time affiliate network APIs (e.g., Amazon Associates, CJ, ShareASale) with dynamic pricing updates.
Autonomous Agent Price Tracking: Enable background agent monitoring where users can instruct their local assistant to alert them when a specific product drops below a target price threshold.
Decentralized Attribution: Explore cryptographic verification and transparent commission tracking for agent-referred conversions.