The document tools are what lift it above a data wrapper: you can ask what management actually said about margins last quarter and get an answer grounded in the transcript, with the index built and cached on your own machine. Two-thirds of it works without any login — only the screening queries need a Screener.in account.
An MCP server that turns an assistant into an Indian equity research desk. Company fundamentals come from Screener.in, corporate events and bulk deals from NSE, and a local RAG pipeline lets you ask questions across annual report PDFs and earnings call transcripts.
- search_company, get_company_overview and get_full_analysis — find a company by name or symbol, read key ratios, price and 52-week range, or pull everything at once
- get_financials, get_quarterly_results and get_shareholding_pattern — P&L, balance sheet, cash flow and ratios; the last eight quarters; and the promoter, FII and DII holding trend
- get_peer_comparison, compare_companies and compare_stocks_ui — the sector peer table, two to five stocks side by side, and an interactive dashboard on hosts that render it
- analyze_red_flags and explain_for_beginners — structured red-flag detection, and a plain-language explainer of what the company does
- screen_stocks, screen_by_theme and list_investment_themes — a custom Screener.in query, or one of the pre-built themes such as high ROCE with low debt or small caps with strong growth
- get_document_list, analyze_annual_report and analyze_earnings_call — the available annual reports and transcripts, and free-text questions answered over their contents
- get_company_announcements and search_shareholder — NSE corporate announcements with category filters, and investor activity via bulk deals
- get_commodity_prices and notebook_ai — commodity price context with the companies it affects, and local research notes you can save, read and summarise
Uv, then claude mcp add screener -s user -- uvx screener-mcp. Company financials work with no login; stock screening needs a free Screener.in account, with SCREENER_USERNAME and SCREENER_PASSWORD exported in your shell. Document analysis needs the extra dependencies (pip install pdfplumber sentence-transformers chromadb, or the [ai] extra) — PDFs are parsed and indexed locally and cached on disk, so a report is never re-downloaded or re-indexed.
One command plus a key — claude mcp add screener -s user -- uvx screener-mcp, then supply credentials