Search works and takes real parameters — price bounds, pagination, sort — which is enough to make comparison shopping a conversation rather than a dozen tabs. The purchase path is the interesting claim and the honest caveat: the README lists cart, checkout and order tracking as coming soon, so treat this as a discovery tool with a payment story attached rather than an end-to-end one.
An MCP server that connects an assistant to SearchAgora, a universal product search engine that discovers products from across the web. Today it covers product search; the cart and checkout side is listed by the author as coming soon.
- Product search across the web from inside a conversation — ask for a category and get results back without leaving the assistant
- Search you can narrow: a query string, results per page, page navigation, a minimum and maximum price, and a custom sort order
- Results you can compare in the chat, so "headphones under $100 sorted by highest rating" is a single request rather than a browsing session
- Cart management and purchase are named in the README as not yet available — add to cart, view cart, checkout, saved favourites and order tracking are all listed under coming soon
No account is needed for search itself. Install [UV](https://docs.astral.sh/uv/getting-started/installation/) and point your client at `uvx agora-mcp` — the package on PyPI is `agora-mcp`. Paying for anything additionally needs an L402-compatible client such as Fewsats, configured alongside as a second MCP server with `FEWSATS_API_KEY` from your Fewsats API keys page. For local development, run it with `uv --directory /path/to/your/agora-mcp run agora-mcp`.
One command — uvx agora-mcp