Connect an AI Assistant to eBay's Official Read-Only API Tool
What it is
Why it matters now
How to try it this week
- Sign up for a free eBay Developers Program account at developer.ebay.com and create an application to get your keys. Check the current programme terms before you start.
- Start in Application Token mode, which uses only your app's Client ID and Secret and gives access to public data. Leave the user-token mode, which reaches your own orders and inventory, until you're comfortable.
- Follow the readme to add the server to an MCP-capable assistant. It runs on your own computer with Node.js.
- Project idea, the item-specifics gap check: pick a category you sell in a lot and ask the assistant to use the Taxonomy API (getItemAspectsForCategory) to list the required and recommended item specifics. Compare that list with a few of your own live listings and note what's missing.
- Second idea: use the Browse API to look at how active listings for one product are titled and priced right now. That's live listings only, not sold data.
Human-in-the-loop checkpoint
Cost and energy note
Security notes
- Your Client Secret and any refresh token are like passwords. Keep them in the tool's config on your own machine. Never paste them into a chat, a forum post or a shared document.
- Stick to the official eBay repository. Several unofficial "eBay MCP" copies exist on GitHub. Some are genuine community projects, but they can have much wider write access. Read before you install.
- Revoke and regenerate keys if you think they've been exposed.
Pitfalls
- Marketplace Insights, the sold-data API, is limited-access, so don't plan a project around it unless eBay has approved you.
- Make sure you're querying the UK marketplace (EBAY_GB) and not the US default.
Sources: github.com/eBay/npm-public-api-mcp | developer.ebay.com/api-docs/commerce/taxonomy/resources/cat… | developer.ebay.com/api-docs/buy/browse/overview.html
#aiprojects #api
- Question
- Active discussion
- Answer emerging
- Community solution
- Knowledge article