Indian Market Intelligence

Shared data components for an Indian equity market intelligence system.

This slice is a canonical company registry plus a read-only MCP server over Streamable HTTP. There is no FastAPI app, database, live market feed, or authentication yet.

Layout

• data/companies.json — one object per listed company

• src/company_registry.py — read-only lookups over that file

• src/mcp_server.py — MCP tools over the same registry

• tests/ — registry and MCP tool tests

Related MCP server: screener-mcp-pro

Company fields

Unknown values are stored as null. Do not invent identifiers, websites, or social handles.

The seed set is Indian defence / aerospace names (plus L&T, which has a large defence business). ISINs reflect post-split equity where listing pages show a newer code.

Python usage

from company_registry import (
    get_all_official_x_handles,
    get_companies_by_sector,
    get_company_by_name,
    get_company_by_symbol,
    get_defence_research_universe,
    get_official_x_handle,
    get_official_x_handles_by_sector,
    search_companies,
)

get_company_by_symbol("hal")
get_company_by_name("MDL")
search_companies("ship")
get_companies_by_sector("Aerospace & Defence")
get_official_x_handle("BEL")
get_all_official_x_handles(sector="Ship Building")
get_official_x_handles_by_sector("Ship Building")
get_defence_research_universe()

Lookups are case-insensitive. get_company_by_name and search_companies both honour aliases.

MCP server

The server uses the official Python MCP SDK (mcp) and Streamable HTTP (not the legacy SSE transport). Tools are read-only wrappers around company_registry.py.

It binds to 0.0.0.0 and listens on PORT when that environment variable is set (Railway sets this automatically). If PORT is unset, it uses 3001 for local development.

Endpoints

A browser can open / or /mcp to confirm the server is up. /mcp is still the MCP endpoint; Chrome just gets a status page instead of hanging on the event stream.

Local start

This machine has python3, not python:

python3 -m pip install -r requirements.txt
python3 src/mcp_server.py

That listens on:

• Health: http://127.0.0.1:3001/health

• MCP: http://127.0.0.1:3001/mcp

Cursor is configured in .cursor/mcp.json to use http://127.0.0.1:3001/mcp. Start this server first, then reload MCP in Cursor.

To mimic Railway locally:

PORT=8080 python3 src/mcp_server.py

Railway deployment

1. Create a new Railway project from this repo (GitHub, or railway up).

2. Railway installs requirements.txt and starts the web process from Procfile / railway.toml.

3. Generate a public domain in Railway (Settings → Networking).

Start command (also in Procfile):

python src/mcp_server.py

After deploy:

• Health: https://<your-railway-domain>/health

• MCP: https://<your-railway-domain>/mcp

Set Railway’s health check path to /health if you want deploy probes against that endpoint.

Environment variables

No other environment variables are required. There is no database, API key, or auth config.

Tools:

Tests

python3 -m pip install -r requirements.txt
PYTHONPATH=src python3 -m pytest tests