Stock Screener MCP: Screen NSE and BSE Stocks with AI

A stock screener MCP lets you screen NSE stocks with AI by describing what you want in plain English. Tapetide runs a screener MCP server that connects Claude, ChatGPT, Cursor and other MCP clients to screening tools covering about 8,200 NSE and BSE listed companies, including small caps and SME stocks. You ask for "profitable mid caps with low debt and rising sales", the assistant picks the filters, runs the screen and explains the shortlist. This guide covers the screening tools, how to iterate a screen in conversation, six example prompts and the limits you should know.

If you prefer clicking filters, the same data powers the web screener. The MCP route is for when you want to reason about the results, not just list them.

#How the screener MCP server works

The Tapetide MCP server is a remote endpoint at https://mcp.tapetide.com/mcp. Once connected, your assistant can call more than 50 tools. Four of them handle screening.

#screen_stocks: 326 fundamental ratios

This is the main fundamental screener. It filters on 326 ratios across valuation (P/E, P/B, EV/EBITDA), profitability (ROE, ROCE, margins), growth (sales and profit growth over several periods), balance sheet strength (debt to equity, interest coverage), dividends and market cap. You describe the conditions; the assistant translates them into filters and returns matching stocks with the ratio values.

#screen_stocks_technical: RSI, MACD, moving averages, volume

For price behaviour, screen_stocks_technical filters on RSI, MACD, moving average positions and crossovers, and volume. Useful for questions like "which Nifty 500 stocks are above their 200-day average with RSI under 40".

#get_screener_ratios: discover filter names

Ratio names have to match exactly. get_screener_ratios lists the available fields, so the assistant can check that "return on capital employed" maps to the right one before it runs a screen. If a result looks off, asking the assistant to call this tool first usually fixes it.

#run_preset_screen: ready-made screens

run_preset_screen runs predefined screens without you specifying every condition. It is a quick starting point that you can then narrow in conversation.

Related tools help you go deeper on what a screen returns: get_company_profile, get_financials, get_shareholding, get_tapetide_score (see the score methodology) and screen_tapetide_scores for screening on the score itself.

#Connect the screener MCP to your AI

Most clients need only the URL and a Google sign-in:

• Claude.ai: setup guide
• ChatGPT: setup guide
• Claude Code: setup guide
• Cursor: setup guide
• VS Code: setup guide

For editors and agents that read an mcpServers config, this is the entry:

{
  "mcpServers": {
    "tapetide": {
      "url": "https://mcp.tapetide.com/mcp"
    }
  }
}

The full client list is on the MCP overview.

#Iterate a screen in conversation

The advantage over a filter form is that the screen becomes a dialogue. A typical session runs in four moves.

1. Start broad. "Find NSE stocks with ROE above 18% and debt to equity below 0.5." You might get 150 names.
2. Narrow. "Keep only those with market cap above 5,000 crore and 3-year sales growth above 12%." The assistant reruns screen_stocks with the extra filters.
3. Add a different kind of condition. "Of these, which are trading above their 200-day moving average?" Now it calls screen_stocks_technical and intersects the lists.
4. Explain the shortlist. "For the top five, summarise what drives the ROE and flag anything unusual in shareholding." The assistant pulls get_financials and get_shareholding for each and writes a short note per stock.

A few habits make this work better:

• Name the units. "Market cap above 5,000 crore" is clearer than "mid cap", which the assistant has to interpret.
• Ask it to show the filters it used. You can then correct a threshold rather than restating the whole screen.
• Ask for a table with the ratio values, not just names, so you can see how close each stock is to the cut-off.
• Keep screens tight. Each rerun is a tool call, and the free plan allows 50 per day.

#Six example screening prompts

Copy these into any connected client and adjust the thresholds.

Quality

Screen NSE and BSE stocks with ROCE above 20% for each of the last 3 years,
debt to equity below 0.3 and positive operating cash flow. Show a table with
ROCE, debt to equity and market cap, sorted by ROCE. Include the observation date.

Value

Find stocks with P/E below 15, P/B below 2 and ROE above 15%, excluding
companies with market cap under 1,000 crore. Call get_screener_ratios first
to confirm the field names, then list the filters you used.

Growth

Screen for companies with 3-year sales growth and 3-year profit growth both
above 20%, operating margin above 15%. For the top 10 by profit growth,
summarise the last four quarters from get_financials.

Dividend

Find stocks with dividend yield above 3%, a payout ratio below 70% and
interest coverage above 5. Flag any where profit fell in the latest year.

Momentum

Use screen_stocks_technical to find stocks above their 50-day and 200-day
moving averages with RSI between 55 and 70 and volume above the 20-day
average. Then remove any with debt to equity above 1.

Small cap

Screen small caps with market cap between 500 and 3,000 crore, ROE above 15%,
promoter holding above 50% and no increase in promoter pledge. Show
shareholding for the top five and state the date of each figure.

#Save results to a watchlist

When a shortlist looks worth tracking, ask the assistant to keep it:

Add the top eight symbols from this screen to my Tapetide watchlist,
then show me the updated watchlist.

The assistant calls add_to_watchlist and then get_watchlist. You can remove names later with remove_from_watchlist, or move stocks you actually hold into your research portfolio with add_portfolio_stocks. These tools edit only your Tapetide research records. Nothing places orders at a broker.

Saved stocks show up on the site too, so you can open any of them on its stock page, for example /stocks/RELIANCE, to check charts and filings.

#Caveats

Screening with AI is fast, but it inherits the limits of both the data and the model.

• Not live data. Prices are the latest available, not a trading feed. Ratios follow each company's reporting period, so a screen run today may use last quarter's numbers for one company and an older period for another. Ask for the observation date.
• Filter interpretation. The assistant translates your words into filters. Vague terms like "cheap" or "strong balance sheet" can be mapped in different ways. Ask it to list the exact filters before you trust a result.
• Sector context. A debt to equity cut-off that suits manufacturers will exclude most banks and NBFCs. Screen financials separately or ask the assistant to treat them differently.
• Survivorship and one-offs. A single exceptional year can push a company through a growth or ROE filter. Multi-year conditions reduce this.
• Tool call budget. Iterating a screen uses calls. The free plan includes 50 successful calls per day and up to 1,000 per month; see pricing for higher limits.
• Model mistakes. Assistants can misread a table or merge two results. Spot check numbers on the stock page.

The data is aggregated from multiple licensed and public sources and structured by Tapetide. Treat every screen as research input, not a recommendation to buy or sell.

#Where to go next

Pick your client from the MCP overview and connect in a couple of minutes. If you use Claude, start with the Claude.ai setup. Then try the quality prompt above, narrow it twice, and save what survives to your watchlist.