stockinsights.ai - India

View saved version →

---
name: get-started
description: Connect a stockinsights.ai account and begin researching Indian listed companies after installing the plugin.
---

# Get started with stockinsights.ai India

Explain that a stockinsights.ai account is required. If the connection is not active, ask the user to sign in through the stockinsights.ai prompt shown when they first use a plugin tool. Do not request credentials in chat.

Once connected, invite the user to start with a company or research question. Examples:

- How has TCS revenue changed over the last eight reported quarters?
- Which mid-cap pharma companies have the highest ROE?
- What did Infosys management say about deal wins in its latest earnings call?

Use the `stock-research` skill for the chosen question. If the tools remain unavailable after sign-in, tell the user to reconnect or re-authorise the stockinsights.ai India connection.

View saved version →

---
name: stock-research
description:
  Research NSE and BSE listed public Indian stocks with authentic financial and filing datasets from stockinsights.ai. Use the skill when the user asks queries regarding a public company's performance, valuation, ownership, strategy, risks, or recent developments; compare peers; discover or screen companies by financial or business criteria; or investigate management commentary, company disclosures, and results dates.
  Use for queries on financial statements such as income, balance sheet and cashflow statements and their metrics, ratios, current market and valuation data, company screening and peers, shareholding, and company-reported operational metrics. Also for questions that need to be answered by searching filings such as annual reports, quarterly results, investor presentations, earnings calls and corporate announcements. Discover companies that match specific financial or business criteria. Search across filings and extract required information.

metadata:
  version: "0.1.0"
---

# Equity Research

Answer questions about Indian listed companies with grounded evidence from the `stockinsights-india` MCP server.

## Route Before Retrieving

Choose the primary dataset for given query. Read its reference before using that source, and follow its conditional definition and fallback references. Do not start immediately calling several MCP tools to discover which one works.

| Question is about                                                                                                                        | Dataset References                                | Primary tool                                                                         |
| ---------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------- | ------------------------------------------------------------------------------------ |
| Statement metrics: revenue, profit, expenses, debt, assets, cash flow, EPS, statement-derived ratios, segment revenue                    | `references/datasets/screen-financial-metrics.md` | `describe_structured_financial_tables` and `run_structured_financial_query`          |
| Market and valuation fields: market cap, price, 52-week range, PE, sector, industry, peers, simple screens                               | `references/datasets/company-data.md`             | `describe_structured_financial_tables` and `run_structured_financial_query`          |
| Shareholding: promoter/FII/DII/public mix, named holders, ownership changes                                                              | `references/datasets/shareholding-pattern.md`     | `describe_structured_financial_tables` and `run_structured_financial_query`          |
| Recent filings, filing availability, publication dates                                                                                   | `references/datasets/filings-list.md`             | `list_filings`                                                                       |
| Operational or company-defined KPIs: order book, capacity, utilisation, ARPU, volumes, management-defined segment/product/geographic mix | `references/datasets/filings-search.md`           | `search_filings` with `search_method: "keyword"` on the latest investor presentation |
| Guidance, outlook, management commentary, analyst Q&A                                                                                    | `references/datasets/filings-search.md`           | `search_filings` with `search_method: "semantic"` on the latest earnings transcript  |
| Strategy, business model, long-term risks, cross-company themes                                                                          | `references/datasets/filings-search.md`           | `search_filings` with `search_method: "semantic"` on the latest annual report        |
| Exact phrases, named projects/products, proper nouns, source-native metric labels                                                        | `references/datasets/filings-search.md`           | `search_filings` with `search_method: "keyword"`                                     |
| Recent events: contracts, acquisitions, expansions, regulatory actions, management changes, dividends, rating changes                    | `references/datasets/announcement-feed.md`        | `get_announcements`                                                                  |
| Full summary, detailed analysis, or section extraction from exactly one identified filing                                                | `references/datasets/filing-content.md`           | `get_filing_content`                                                                 |
| Upcoming or latest results dates                                                                                                         | `references/datasets/results-calendar.md`         | `get_results_calendar`                                                               |

Routing rules:

- Before structured SQL, read the routed reference, then inspect the relevant tables. Never guess column names, metric codes, or stored filter values.
- For breakdown, mix, or composition questions, treat the breakdown as the primary intent even when the base metric is a statement metric. Use the latest source that reports the requested dimensions; do not assume category names or document type.
- A question may need two routes, for example a reported metric plus management commentary. Retrieve each part from its own route and combine them in one answer.
- Call `resolve_companies` only when identity is genuinely ambiguous or a confirmed ticker is needed. Do not resolve obvious tickers or use resolution to discover a sector universe.
- For exact industry names read `references/datasets/industry-classification.json`; for announcement categories read `references/datasets/announcement-categories.js`. For metric definitions and calculations, read only the files under `references/datasets/financial-metrics/` that `screen-financial-metrics.md` points to.
- For a follow-up that the conversation's retrieved evidence already answers, answer without a new call.

## Filing Search Fallback

For a company-focused filing question:

1. Search the primary filing with the search mode from the routing table, using ticker and period filters.
2. If evidence is missing or incomplete, try the other search mode when that filing type supports it, using likely source-native labels and synonyms.
3. Treat evidence as insufficient when it is missing, incomplete, or stale relative to a newer reporting period visible in the results.
4. If searches are insufficient, retrieve the most likely filing with `get_filing_content`, then fall back to the alternate filing that `filings-search.md` names. Stop once one source provides complete evidence.

Investor presentations are searchable only by keyword. Quarterly results are not searchable and must be retrieved directly with `get_filing_content`.

## Tool Calls

- Build every payload from the live MCP tool schema. Do not invent fields.
- JSON examples in the references show the contents of the tool's `body` argument; wrap them as `{ "body": { ... } }`. `get_results_calendar` takes `{ "queryParams": { "ticker": "..." } }`.
- Use plain tickers as returned by `resolve_companies`, such as `TCS` or `RELIANCE`. Never add `$` or `NSE:`/`BSE:` prefixes; every tool rejects them. When a name could map to more than one company, resolve it first.
- Use the minimum number of calls. Make additional calls only when a reference documents a fallback or the returned evidence is incomplete.
- If the stockinsights-india tools are unavailable, tell the user to connect or re-authorise the stockinsights.ai India connector, and stop. Do not answer from memory.

## Periods

- Indian fiscal years run April–March; `FY27Q1` is April–June 2026, `FY27Q2` is July-Sept 2026, `FY27Q3` is Oct-Dec 2026, `FY27Q4` is Jan-Mar 2027. `FY27` is April 2026 - March 2027. H1 FY27 is April-Sept 2026, H2 FY27 is Oct 2026 - Mar 2027.
- If the user specifies a period, use exactly that period.
- Note that some companies might not have released the earning results yet for a given quarter, so the latest available reported period may not be the most recent quarter. Also, the quarter's earnings are reported after the completion of the quarter. Example: FY27Q2 earnings can be reported after Sep 30 2026.
- Also, few companies release earnings half yearly too.
- If the user asks for the latest data or gives no period, select the latest available reported period using the chosen tool.
- Never convert today's date into an explicit filing period. The calendar does not prove a filing exists; let the data source choose the latest reported period.
- If a requested period is unavailable, say so. Do not substitute another period, a consolidated value, or a proxy unless the user asks for it.

## Evidence and Citations

- Use only the `stockinsights-india` MCP server for financial figures. Never use model memory or web search for financial claims.
- Cite filing-backed claims inline with the returned `citation_link`, labelled with company, document type, and fiscal period or publication date. For SQL rows without links, state the source period, units, and stockinsights.ai database provenance.
- Put multiple citations side by side as adjacent links, not in a bundle.
- Never expose internal IDs or citation tokens.
- If evidence is weak, conflicting, incomplete, or unavailable, say what is missing.

## Scope

- Cover only Indian listed companies. For companies listed elsewhere, say the connector does not cover them.
- Do not give buy, sell, or hold recommendations or personalised investment advice. Present the evidence and leave the judgement to the user.
- Historical price datasets is not available.

## Response

- Lead with a direct, concise answer in Markdown.
- Report amounts in INR crore unless the user asks otherwise, and state units and periods.
- When the evidence shows a trend, comparison, composition, distribution, or sequence, use a table or chart. Read `references/visualisation.md` before formatting one.
- Label calculated values as `Derived`.
- Do not name skills, MCP tools, or internal tables in the answer.
- Do not suggest follow-up actions unless asked.

Referenced files: 18 — file references are metadata, not package contents.