DalalOS

MCP Server for Indian Stock Research

GitHub

Documentation

DalalOS — Indian stock research for AI agents

DalalOS is a hosted Model Context Protocol (MCP) data layer for researching Indian NSE/BSE-listed companies with AI assistants and agents.

What DalalOS provides

DalalOS exposes structured Indian-market research data to MCP-compatible clients, including:

  • company and security search;
  • end-of-day quotes and price history;
  • company financial statements and mechanically-computed ratios;
  • peer and sector comparison data;
  • shareholding and promoter-pledge data;
  • corporate actions, results, filings and disclosed events;
  • screening, indices and market context;
  • official macro/reference datasets used for research context; and
  • an account-scoped DalalOS watchlist.

Data is sourced from official and public Indian-market sources such as NSE archives, BSE official interfaces and exchange filings, SEBI-mandated BSE-XBRL filings, and the named official public-sector publisher for supported macro/reference datasets.

Important boundaries

  • Not real-time: quote-style stock-market data is end-of-day, not a live tick or order-book feed.
  • Research, not execution: DalalOS is not a broker and does not place market orders or execute trades.
  • No investment advice: DalalOS returns sourced facts and mechanically-computed measures, not buy/sell calls, target prices or investment recommendations.
  • Read/write scope: market-data research tools are read-only. The explicit write exception is the caller's own DalalOS watchlist, where authenticated users can add or remove symbols.
  • Missing data stays explicit: unsupported, unavailable or stale coverage is surfaced rather than replaced with invented values.

Connect

Use the hosted streamable-HTTP endpoint:

https://mcp.dalalos.in/mcp

OAuth is the preferred authentication path for clients that support it. API-key Bearer authentication is available as a fallback for compatible clients that require it.

See docs/CONNECT.md for public-safe setup notes and the maintained website guides for client-specific instructions.

Public capability contract

capabilities.json is the deliberately minimal machine-readable discovery contract for the hosted MCP surface. It contains only public MCP tool names plus read/write, destructive and deprecation lifecycle facts, along with derived registered/preferred/write counts.

It intentionally does not publish tool implementation modules, input-schema hashes, REST mappings, storage/cache details, source-adapter logic, refresh mechanics, deployment information, credentials, or other backend internals.

Deprecated compatibility aliases remain in the registered-name list but are excluded from preferred_tool_count, so directories and product surfaces can avoid inflating capability counts with old aliases.

MCP Registry identity

The public registry manifest is server.json. DalalOS already has an official MCP Registry identity under:

io.github.aatharva16/dalalos

That existing identity is intentionally preserved so the hosted endpoint has one canonical Registry record. Registry metadata changes must be published using credentials authorized for that existing namespace; this public repository does not store those credentials.

Repository purpose

This repository is the public project, integration and discovery surface for DalalOS. It exists so users, MCP registries, AI-tool directories and open-source ecosystem lists have a stable public GitHub reference for the hosted service.

It intentionally does not contain the DalalOS backend implementation. In particular, this repository does not publish internal ingestion logic, database/cache schemas, source-adapter code, deployment topology, credentials, private operational tooling, private issue/PR history, or unpublished architecture.

Public files in this repository are limited to integration metadata, connection documentation, public capability/provenance descriptions and safe usage examples.

Public metadata

Legal and privacy

DalalOS is not affiliated with or endorsed by NSE, BSE, SEBI, OpenAI, Anthropic, Cursor, or another AI-client vendor unless explicitly stated otherwise by that party.