Merchant’s Assistant

Inspiration

When I reviewed the hackathon details, I started thinking about what I would really want to do with an AI agent. Usually, doing a task purely on your own is faster than delegating it, or having the agent do it completely autonomously is faster than doing it yourself. I wanted to focus on the type of task where a human and an agent work together efficiently.

Looking at the sponsors, I saw Shopify. Having done some research into dropshipping and e-commerce, I know how much work goes into managing a store. Fully autonomous agents run into permission and safety issues, since an agent cannot just be allowed to do everything without supervision. In physical retail stores, owners have managers and assistants to help run day-to-day operations. For an online store, an agent can play that exact role. That is where the name Merchant Assistant comes from: an assistant to the merchant who owns the store.

What Merchant Assistant Does:

1. Product Research and Proposals: The task that takes merchants the longest is research. When prompted by the merchant, the agent researches top-performing products in specific categories using ShopifyQL under the hood, exposed via WebMCP. Instead of just pulling data, the agent creates actionable proposals for items the merchant can procure and add to the store.
2. Interactive Control via ChatGPT: Through WebMCP, merchants can interact with the agent directly in the ChatGPT app or on the web. Once the agent creates a proposal, the merchant reviews and accepts it. The changes can then be applied immediately with a single click, or the merchant can instruct the agent to apply multiple approved proposals.
3. Product Audits and Optimization: The agent audits existing listings, checking descriptions and top-performing products based on sales data. It highlights areas for improvement and suggests updated descriptions aligned with current market trends.

The merchant remains in control: they trigger the agent, review its findings, and approve every change.

How I Built It:

I built the project using WebMCP, ShopifyQL, and Shopify MCP to give the agent the tools needed to query store analytics and execute store actions.

Challenges I Ran Into:

1. Shopify Permissions: Securing Level 2 permissions to read merchant reports and sales data, which are necessary for the agent to audit the store and propose optimizations.
2. Judge Access and Testing: Shopify's standard security makes it difficult to grant external test access simply via email. I implemented a workaround so judges can test the integration live.

What I AM Most Proud Of:

The entire end-to-end system is 100% functional and working live.

Built With

• javascript
• react
• shopifyql
• webmcp