Positioning & One-Page Conclusion

0. One-page conclusion (TL;DR)

• The opportunity is real. In 2026 MCP (Model Context Protocol) is the de-facto standard for AI Agents to connect to tools, and supply has exploded: the official Registry shows 1,600+ unique Servers after dedup (within the sampled range), the Smithery public API reports 19,932, the PulseMCP sitemap contains 21,710 server pages, and npm lists 9,775 packages under the mcp-server keyword. But the "who lists the most" layer is already a red ocean.
• The gap is clear. Existing directories all compete on coverage; nobody turns "does this Server actually work right now, on which client does it run, does it cost money, how is auth configured, when was it last down" into a trustworthy, subscribable, licensable data product. Glama has already placed a commercial Data License on its API data (key required, unauthorized redistribution forbidden) — which is itself the strongest market signal that structured MCP data commands payment and can carry a licensing moat.
• The current MCPera site is a pure static landing page: hard-coded cards, fake statistics (1,200+ / 300+ / 50+), a placeholder search, and a single sample detail page. It holds the good name mcpera.com and a "vertical MCP service" position but has no real data asset. This is exactly the gap identified by card A-006: upgrade a static yellow-pages site into a verified MCP supply-chain database.
• Recommended play (narrow but true, startable now): do not build a "whole-industry encyclopedia"; build a Verified MCP Supply-Chain DB —
  1. use the official Registry + Smithery + npm + GitHub as the compliant, traceable free cold-start base;
  2. layer on a proprietary verification layer others skip (liveness probing, client compatibility matrix, pricing/auth extraction, time snapshots);
  3. use programmatic SEO to turn every structured record into a long-tail landing page (/server/<slug>, /<category>/, /vs/a-vs-b/, /pricing/);
  4. monetize via B2B data licensing + premium subscriptions + referral commissions, not a bet on consumer ads.
• Feasibility conclusion: technically it can be started by one person + scripts (data sources are public, the MCP protocol can be handshaked programmatically, SQLite + static JSON comfortably supports tens of thousands of records). The real moat is not the crawler but the three dirty jobs — continuous verification + historical time series + data-licensing compliance — which is precisely the consensus across the five 777-007 cards.

---

1. Background & problem definition

1.1 State of the MCP ecosystem (measured 2026-10)

MCP is an open protocol led by Anthropic and now broadly adopted by mainstream Agent clients (Claude, Cursor, and various IDE/Agent frameworks) for exposing external tools/data sources to an LLM in a standardized way. Supply splits into two types:

• Local / package type: distributed via npm/pip/Docker and self-hosted by the user. Of the 5,200 sampled version records in the official Registry, 2,109 carried packages (local distribution).
• Remote / hosted type: serves an endpoint over streamable-http / SSE, operated by a vendor and often tied to billing. In the same sample, 4,228 carried remotes — remote-ization and servitization are already mainstream, and this is precisely the layer most fertile for commercialization.

Measured scale (2026-10-13; raw commands and counting rules in Appendix A):

Data source	Measured metric	Value	Nature of data
Official Registry registry.modelcontextprotocol.io	sampled version records / unique Servers after dedup	5,200 / 1,673 (reached the ai.starreel prefix; not a full crawl)	official metadata incl. remotes/packages/status/version
Smithery public API registry.smithery.ai	pagination.totalCount	19,932	includes useCount/verified/remote/createdAt/score
PulseMCP sitemap	server detail-page URLs	21,710	SEO indexing scale
npm search	packages with keyword mcp-server	9,775	code-package supply
Docker Hub search	mcp server images	408,334 (noisy; order-of-magnitude only)	image supply, high noise
GitHub modelcontextprotocol/servers	stars	91,206	traction of official reference implementations

Implication: supply is already past ten thousand and inflating fast, and discovery/trust/selection costs are deteriorating sharply — the fundamental precondition for a directory-type data business.

1.2 Core pain points (segmented by user; detail in chapter 3)

1. Developer selection pain: among thousands/tens of thousands of Servers, descriptions are all self-promotional; only after cloning do you find it won't run / is deprecated / needs a private key / is incompatible with your client version. The official Registry explicitly carries status: deprecated (67 in the sample), but most directories don't emphasize alive-vs-dead.
2. Enterprise procurement pain: when a company wants to integrate a paid MCP (search, financial data, CRM, legal data), it cannot quickly answer "who is reliable, how do they bill, is auth secure, what's the SLA, how costly is switching". There is no Gartner-style comparable matrix.
3. MCP vendor customer-acquisition pain: tens of thousands of Servers are heavily commoditized (in the sample alone the single name ai.bowmark had 10+ version numbers); vendors lack a neutral distribution/endorsement channel that brings precise traffic and a trusted badge.
4. Agent-platform integration pain: Cursor-like clients and Agent orchestration platforms want to embed a "tool marketplace" and need a clean, licensed, compatibility-annotated, programmatically callable vendor library rather than scraping a pile of HTML.

1.3 Why MCPera (fit with the existing position)

• The domain and brand are the category itself (MCP + era); it's currently occupied but assetless, so opportunity cost is minimal;
• the existing site already targets a global audience in English and has an SEO foundation (canonical, WebSite/SoftwareApplication JSON-LD, semantic headings), so it can directly host programmatic long-tail pages;
• the five 777-007 cards complete the methodology: A-006 sets the direction (Registry database), A-003/A-013 define the moat (real-test verification matrix + version/snapshot + source URL / verification date / staleness flags), A-015 defines cold-start (self-produced + public structured sources + UGC, compliant and traceable), and A-007 defines monetization and pragmatic execution.