DalalOS: Indian Stock Market Research (NSE/BSE)

Indian stock market (NSE/BSE) MCP server

DalalOS is a hosted Indian stock market (NSE/BSE) MCP server. It gives Claude, ChatGPT, Cursor, Claude Code, Windsurf and other MCP clients 59 tools for company and market data: quotes, financials and ratios, shareholding and promoter pledge, corporate filings (facts with the source page and verbatim quote), order book and other operating metrics for the companies that disclose and are indexed, screens, FII/DII flows, F&O positioning, IPOs, corporate actions and macro series. Each response carries its source and a freshness date. It returns reported facts and mechanically-computed ratios, never a buy/sell call.

You do not need a brokerage account, a data-terminal subscription or an API key. Add the URL to your client, sign in once with Google or an emailed code, and you are on the free tier.

• MCP endpoint (streamable HTTP): https://mcp.dalalos.in/mcp
• Website: https://dalalos.in
• MCP Registry name: in.dalalos/mcp
• PyPI: dalalos-mcp (connection helper and stdio proxy)
• Free to start, no card: https://dalalos.in/pricing

Not investment advice. DalalOS provides market data and mechanically-computed measures from public filings and exchanges. It does not issue recommendations, target prices or trading signals, and it is not a SEBI-registered investment advisor. See boundaries.

Connect

DalalOS uses OAuth: add the URL, sign in once in the browser window your client opens, and you are on the free tier. No API key is needed for the clients below. The server URL is always:

https://mcp.dalalos.in/mcp

Claude (claude.ai and Claude Desktop)

Open Settings → Connectors → Add custom connector, paste the URL above, choose Connect, then complete the DalalOS sign-in window and start a new chat.

Claude Code

claude mcp add --transport http dalalos https://mcp.dalalos.in/mcp

Then run /mcp inside Claude Code to finish the browser sign-in.

ChatGPT

The easiest way is the published plugin: open DalalOS in ChatGPT, select Install plugin, and sign in to DalalOS when prompted (Google, or a code we email you). It is published for ChatGPT users in India.

If the plugin is not available to you, add the server manually. ChatGPT does not use a JSON config file for this: in ChatGPT open Plugins → Create Plugin, choose Server URL, enter https://mcp.dalalos.in/mcp, select OAuth, create the plugin and complete the DalalOS sign-in window. Then start a new chat and select the plugin.

Cursor

Add to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global), then choose OAuth when Cursor offers it:

{

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

VS Code

Add to .vscode/mcp.json:

{

"servers": { "dalalos": { "type": "http", "url": "https://mcp.dalalos.in/mcp" } } }

Codex

codex mcp add dalalos --url https://mcp.dalalos.in/mcp

codex mcp login dalalos

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json, restart Windsurf, then complete the DalalOS sign-in when it opens:

{

"mcpServers": { "dalalos": { "serverUrl": "https://mcp.dalalos.in/mcp" } } }

Windsurf names the remote-URL field serverUrl, not url.

Cline

Add to cline_mcp_settings.json (Cline's MCP settings). This form uses a DalalOS API key from https://dalalos.in/dashboard/keys:

{

"mcpServers": { "dalalos": { "type": "streamableHttp", "url": "https://mcp.dalalos.in/mcp", "headers": { "Authorization": "Bearer " } } } }

Command-only clients (uvx or pip)

Clients that can only launch a local command, such as the Claude Desktop config file, can run the stdio proxy from the dalalos-mcp package. It forwards to the hosted server and needs an API key from https://dalalos.in/dashboard/keys:

{

"mcpServers": { "dalalos": { "command": "uvx", "args": ["dalalos-mcp", "proxy"], "env": { "DALALOS_API_KEY": " " } } } }

Without uv, run pip install dalalos-mcp and use "command": "dalalos-mcp" with the same args and env. Two helper commands: uvx dalalos-mcp check confirms the hosted endpoint answers, and uvx dalalos-mcp config prints a connection snippet for claude, claude-code, codex, chatgpt, cursor, generic or stdio.

Any other MCP client

Use the client's own schema for a remote streamable-HTTP server. The generic shape is:

{

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

Only if a client cannot complete OAuth and requires a Bearer header, create an API key in your DalalOS account and send Authorization: Bearer . Never commit or share a key. More client guides (Zed, Kiro, OpenCode): https://dalalos.in/connect.

Check it works

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 working connection names the DalalOS tools it called and returns source and freshness metadata.

Example prompts

Each prompt names the DalalOS tools your assistant should reach for. You do not need to name them; the assistant picks them from the tool descriptions.

Prompt pattern for when recency matters: *"Name the DalalOS tools used, show the source, the cached date and any stale or coverage warning. Do not substitute web search for a missing DalalOS value."* More: examples/PROMPTS.md.

What is different

• Official sources only. Data comes from NSE and BSE publications, SEBI-mandated BSE-XBRL filings, NSDL, MOSPI, PPAC and FBIL. DalalOS does not use third-party data-vendor feeds.
• First-party and filing-grounded. Financials come from SEBI-mandated BSE-XBRL filings; exchange filings are extracted into facts that keep their source page and verbatim quotes.
• Research only, no brokerage account. Nothing connects to a broker and nothing places an order. The only write is your own DalalOS watchlist.
• Source and freshness on every response. Each result carries its provenance, a cached_at time and a stale flag. When freshness is the question there is a dedicated get_data_freshness tool.
• Honest gaps. Missing, not-applicable and stale data are surfaced as such, not replaced with an invented value. A valid zero and missing data are different things.
• No verdicts. Raw-as-reported figures or mechanically-computed ratios only: no buy/sell calls, target prices or "top picks".

Tools

DalalOS exposes 59 current MCP tools. The live tools/list returned 61 entries on 29 October 2026 (server version 1.28.1): those 59, a ping helper and one deprecated alias (get_fii_activity, use get_stock_fii_flow). Directory sites that count every name the server lists, such as Glama, show 61. capabilities.json lists 60 (the 59 plus the deprecated alias) and leaves out the ping health check. The maintained reference is https://dalalos.in/tools.

| Watchlist | 3 | `get