Yes, and they should actually be the same product — the MCP server IS the middleware. Here's how I'd think about it:
The product is one thing with two faces:
-
Face 1 (for AI): An MCP server that any AI assistant can query. "What's the best photo editing app under $5?" → your server returns structured data about apps in that category, with features, reviews, pricing, ratings. ChatGPT, Claude, Perplexity, Gemini — any MCP client gets instant access.
-
Face 2 (for developers): A dashboard where app developers submit and manage their listing. Input your app metadata, upload screenshots, connect your App Store/Play Store data. The middleware handles structuring it, optimizing it for AI citation, generating schema markup, and serving it via MCP.
Architecture is actually simple:
App Developer → Dashboard (submit metadata)
↓
Structured Data Store
(app info, reviews, pricing, features)
↓
┌─────────┼─────────┐
↓ ↓ ↓
MCP Server API/Feed Schema.org
(AI agents (future (SEO/crawl
query it) ACP/UCP) for Google)
The data goes in once. It comes out three ways:
- MCP server — AI agents query it in real-time (instant)
- Structured feed — ready to plug into ACP/UCP when those protocols expand to software (future-proof)
- Schema.org markup / embeddable widget — developers add to their own site for crawl-based SEO and AI Overviews
What a developer gets when they sign up:
- Submit app once (or connect App Store Connect / Play Console API to auto-pull metadata)
- AI-optimized listing generated automatically (features rewritten in Q&A format, comparison data structured, review summaries in extractable format)
- Their app is immediately queryable by any AI assistant via MCP
- Dashboard showing: "Your app was recommended X times by ChatGPT, Y times by Perplexity, Z times by Claude" (the monitoring piece everyone else is building — but we include it free because we OWN the data layer)
- Schema.org snippet to paste into their marketing site
MVP scope (what we'd actually build first):
Phase 1 — The MCP server + basic ingest:
- Developer submits app via a simple form (name, description, category, pricing, features, App Store URL)
- We auto-pull ratings, reviews, screenshots from App Store / Play Store APIs
- Structure everything into MCP-servable format
- Deploy MCP server that AI assistants can connect to
- Test with Claude (we already have Claude Code) and ChatGPT (supports MCP since March 2025)
Phase 2 — The developer dashboard:
- Auth, listing management, edit your positioning
- Auto-sync from App Store Connect
- Analytics: how often your app gets queried via MCP
Phase 3 — Feed expansion:
- Generate ACP/UCP-compatible feeds when those protocols support software
- Perplexity Merchant Program integration (submit catalog directly)
- Embeddable comparison widgets for developer sites
What makes this defensible:
The classic chicken-and-egg problem actually works in our favor here. The more apps in the catalog, the more useful the MCP server is for AI agents, which means more queries, which means more value for developers, which means more apps. First mover with a critical mass of listings wins.
Also — we know app marketing. This is our agency domain. We can seed it with our own clients' apps and use our agency playbooks to drive adoption. We're not a random developer building an app directory, we're a marketing agency building the distribution layer we wish existed for our clients.
Rough timeline estimate:
- Phase 1 MVP (MCP + basic ingest): 2-3 weeks if we keep it tight
- Phase 2 (dashboard + analytics): 2-3 weeks after
- Phase 3 (feeds + integrations): ongoing as protocols evolve
Want me to write a proper plan doc for this? I can spec out the data model, MCP server endpoints, and the exact MVP scope. Or if you want to just start building, I can set up the project repo and start with the MCP server.