Use ChatGPT, Cursor & other AI tools for Indian stock research

Why connect your AI tool to DalalOS

Large language models are strong reasoners but weak databases. Ask a raw ChatGPT-style agent for a mid-cap company's latest ROCE and it may guess. DalalOS fixes this by giving the agent 56 current MCP capabilities over streamable HTTP, sourced from official NSE archives, BSE official interfaces and SEBI-mandated BSE-XBRL filings across thousands of companies. Market-data research capabilities are read-only; the 2 account-scoped writes only add or remove symbols from your DalalOS watchlist.

The result: AI-assisted stock research where the data is real and the reasoning is yours plus the model's — with no scraping and no verdicts baked in. Because DalalOS speaks MCP, the same endpoint behaves consistently for Claude, ChatGPT-style agents, Cursor, Perplexity and other MCP-capable AI agents.

Connect ChatGPT with OAuth

This flow has been verified with ChatGPT on the web and a personal Plus account. In ChatGPT, open Plugins, choose Create Plugin, and enter the following details. Interface labels and availability can vary by plan or workspace, but ChatGPT does not use an mcpServers JSON file for this flow.

For a first test, ask: “Use DalalOS to fetch the latest available quote and company profile for Reliance Industries. Tell me which DalalOS tools you called, the source of the data, and its freshness date. Do not use web search.”

A successful response identifies the tools it called and includes source and freshness information. Prices are end-of-day rather than a real-time exchange feed. If the server URL is rejected, confirm that it is exactly https://mcp.dalalos.in/mcp; the /mcp path is required.

Connect Cursor and other OAuth-capable clients

Cursor and other clients that support remote MCP OAuth can use the same hosted endpoint. Add https://mcp.dalalos.in/mcp as a remote MCP server, choose OAuth when prompted, and complete the browser sign-in. Use the client's own configuration format and documentation: config schemas differ between clients.

API-key fallback for config-based clients

Use an API key only when a client cannot complete OAuth or explicitly requires a Bearer header. Sign in to DalalOS, open Dashboard → API keys, and create a key. The example below applies only to clients that accept themcpServers configuration shape.

{
  "mcpServers": {
    "dalal-os": {
      "url": "https://mcp.dalalos.in/mcp",
      "headers": {
        "Authorization": "Bearer <your-isk_live_-key>"
      }
    }
  }
}

Replace <your-isk_live_-key> with the key you copied. Treat it like a password — anyone holding it can query DalalOS on your behalf. Never commit it to a repository or share it in a screenshot.

Generic OAuth configuration

For clients that accept the standard mcpServersshape and handle OAuth automatically, use this keyless configuration. Do not add an API key unless the client asks for one.

{
  "mcpServers": {
    "dalal-os": {
      "url": "https://mcp.dalalos.in/mcp"
    }
  }
}

Example prompts

For developers and builders

Because market-data responses come back in one consistent envelope (status, data, source, freshness), it is easy to compose tool calls into personal AI agents for Indian stock research, dashboards, or research scripts — whether you drive them from Cursor, a CLI agent or your own code. Research capabilities do not place trades; the only write surface is the caller's own DalalOS watchlist, where authenticated add/remove actions save symbols rather than touching a brokerage account.

• Chain tools (search → quote → financials → peers) inside an agent loop.
• Build repeatable research workflows that always cite data freshness.
• Keep the investor framing: the output is data, not advice.

Good practices for AI-assisted research

• Ask the agent to state the data freshness timestamp it used.
• Treat ratios as inputs to your own analysis, not conclusions — DalalOS computes them mechanically and offers no opinion.
• Cross-check anything important against the primary filing before acting on it.
• Remember the data is end-of-day, not intraday.

Ready to put it to work? See using ChatGPT for Indian stock analysis or the broader guide to Indian stock market data for AI agents.