Connect your AI client via MCP

Endpoint (paste into an MCP client — anonymous, free, no key)

https://mcp.koalcheck.com/mcp

Or copy this and send it to your AI to set it up:

Add the "koalcheck" MCP server to my client and connect. It is a remote Streamable-HTTP server at https://mcp.koalcheck.com/mcp — connect anonymously, no API key or login needed (the free tools work right away). If my client supports a remote/HTTP MCP server directly, just use that URL. If my client config only accepts a local command (e.g. Claude Desktop's mcpServers, which rejects a bare "url"), bridge it with command "npx" and args ["-y","mcp-remote","https://mcp.koalcheck.com/mcp"]. Then tell me to fully restart the client, and list the koalcheck tools you can call. It provides US-equity research tools built on an analyst track-record engine (who has actually been right, priced vs SPY, cross-checked with SEC & FINRA).

Capabilities User guide

← @tengyanai

$GLW

BULLISH2026-06-21

Price around the call

↑ 2026-06-21

By horizon✗ miss-27.0%

✗ -6.1%

✓ +22.3%

21d

✗ -27.0%

Original post

Teng Yan · 2026-06-21

AI's next bottleneck is...a glass rod? let me try convincing you 1/ a GPU rack needs roughly 36x more fiber than a standard CPU rack. data centers are expected to swing from under 5% of global fiber demand in 2024 to about 30% by 2027. the marginal growth buyer has shifted. 2/ the preform rod is the real gate. glass fiber is drawn from a preform, a precision glass rod built up layer by layer. draw towers are quick to add. preform capacity takes 18-24 mths per line, before qualification. big Western additions mostly land after 2028. 3/ only ~20 preform producers exist worldwide. the top five hold close to 60% of capacity. YOFC and Corning ($GLW) trade the global #1 spot. Shin-Etsu and Sumitomo hold the specialty edge. the US narrows to one integrated name. 4/ the light-guiding core in

View on X ↗