Daito

MCP server for screening Indian stocks and mutual funds. Wraps screener.in and Morningstar India so you can run fundamental queries straight from Claude or Cursor.

Deployed as a Next.js app on Vercel. The MCP endpoint is at /api/mcp.

Tools

query_india_screener

Runs screener.in's query language against NSE/BSE-listed equities. Same syntax you'd use on the website:

Market Capitalization > 500 AND Price to earning < 15 AND Return on capital employed > 22

Supports 100+ fundamental ratios including annual, quarterly, and multi-year growth variants. Screener.in caps queries at 1500 characters (~40 conditions).

Requires a screener.in account. Ad-hoc queries are login-gated on their end, so you'll need to set SCREENER_IN_EMAIL and SCREENER_IN_PASSWORD. Free account at screener.in/register.

get_india_fund_screener

Screens ~12,700 India-domiciled open-end fund share classes via Morningstar India's API. No credentials needed.

Filters: category (55 Morningstar India categories), star rating, risk rating, expense ratio, AUM, yield, manager tenure, trailing returns (1D to 10Y). Each fund shows up once per share class (Direct/Regular × Growth/IDCW); use term: "Dir Gr" to narrow to direct growth plans. Values in INR, NAVs update daily.

Setup

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "daito": {
      "url": "https://your-deployment.vercel.app/api/mcp"
    }
  }
}

Restart Claude Desktop. For local dev, swap in http://localhost:3000/api/mcp.

Cursor

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json in your project:

{
  "mcpServers": {
    "daito": {
      "url": "https://your-deployment.vercel.app/api/mcp"
    }
  }
}

Restart Cursor. For local dev, use http://localhost:3000/api/mcp.

Running locally

bun install
cp .env.example .env   # fill in screener.in credentials
bun run dev

Other commands:

bun run build      # production build
bun run start      # serve production build
bun run typecheck  # tsc --noEmit

Environment variables

SCREENER_IN_EMAIL and SCREENER_IN_PASSWORD are only needed for query_india_screener. The fund screener works without credentials.

How it works

app/api/[transport]/route.ts  (mcp-handler, Zod validation)
    ↓
lib/services/
    ├── screener-in.ts          → Django CSRF login → /screen/raw/ → cheerio HTML parse
    └── india-fund-screener.ts  → public Morningstar JWT → ecint/v1/screener (FOIND$$ALL)

The screener.in service does a Django CSRF login, caches the session cookie, and re-auths when it expires. The Morningstar service scrapes a public retail JWT from their quickrank page and refreshes it from the JWT exp claim.

Deployment

Hosted on Vercel. vercel.json bumps maxDuration to 60s on the transport route since screener.in can be slow. Node runtime.