What if AI assistants could search real shopping offers without scraping websites or relying on invented discounts?
That's what we've been working on at DiscountHub.
We've launched a public Model Context Protocol (MCP) server that connects compatible AI agents to our existing discounts and promotions catalogue.
The integration
Our server exposes 8 read-only tools:
search_deals — Search product discounts.
find_promo_codes — Discover available coupon codes.
get_store_offers — Get deals and promotions from a merchant.
get_best_deals — Find highly discounted products.
list_stores — Discover stores.
list_categories — Explore product categories.
get_offer_details — Retrieve an offer by ID.
compare_deals — Compare search results.
Transport: Streamable HTTP
MCP endpoint: https://mcp.discounthub.uz/mcp
Testing with a real AI agent
We connected DiscountHub MCP to GitHub Copilot and ran several natural-language tests.
For example, the user asked in Russian:
"Find a laptop with at least 30% off."
Copilot independently selected DiscountHub's tools and returned a real product offer with its listed price, discount and shopping link.
We also tested promotional coupon discovery and scenarios with no matching results.
During testing, we found that regional merchant names could cause incomplete search results. We improved store-name resolution and added regression tests.
Official MCP Registry
DiscountHub is now registered as:
uz.discounthub/deals-and-coupons
Version: 1.0.0
The published server is accessible without requiring an account for public catalogue queries.
Why this matters
We're exploring a future where shopping discovery starts with a natural-language request to an AI assistant.
Rather than building another agent, we're making DiscountHub's catalogue available to the agents people already use.
Documentation: https://discounthub.uz/developers/
Official Registry: https://registry.modelcontextprotocol.io/?q=discounthub
Website: https://discounthub.uz
We're interested in feedback from developers building shopping assistants, commerce agents, and MCP integrations.
Top comments (1)
Nikolas here, co-founder of ApyHub. How did you fix the regional merchant names, an alias table or a fuzzy match at query time? From our experience at ApyHub, a catalogue like yours is easy for agents to use once it sits next to other APIs behind one MCP. We have nothing for deals or promo codes yet, so if you want to list it: around 10 minutes, no exclusivity, apyhub.com/become-a-provider.