AI tools

The AI tools are the actions you can ask the AI Console to perform in plain English. You describe what you want, and the assistant picks the right tool, fetches the data on your behalf, runs the analysis, and returns the result inline. This page explains what the tool catalog covers, how the assistant chooses a tool, how to read the results, and what standard math a tool runs on your own data. The AI Console is available on the Free, Pro Lite, and Pro plans; usage is metered by GNG Credits, and a few heavier actions enforce their own per-tier gate at run time.

Before you begin

Sign in to your account. The tools run inside the AI Console.
To use the portfolio tools, build or connect a portfolio first. See Build your first portfolio.
To delegate to a specialist helper, set one up first. See Subagents and Create a subagent.

What it shows

You do not invoke tools directly. You ask a question, and the assistant chooses and chains the tools it needs. A read-only reference sheet, opened from the console, lists every tool the assistant can call, grouped as in the following sections.

Market data tools

Company search: finds stocks and ETFs by symbol or name, used to confirm a ticker before any ticker-scoped action.
Company profile: the master record for one ticker, including sector, industry, market cap, and dividend information.
Quick price snapshot: current price, the 52-week high and low, recent returns over several windows, and the most recent dividend or split.
News and sentiment: recent news articles for a ticker, each carrying a sentiment score.

Fundamentals tools

Research Terminal metrics: the aggregated Research Terminal data for one ticker, covering price, yield, valuation, profitability, safety, and growth.
Financial statements: the balance sheet, income statement, or cash flow, annual or quarterly.
Earnings history: past earnings (reported earnings per share against estimate, surprise percent) and upcoming earnings dates.
Earnings transcripts: full earnings-call transcripts with quarter, fiscal date, speaker and segment counts, and call-tone readings.
Dividend history: past dividend payments, most recent first.
Insider transactions: executive buying and selling, with name, title, shares, price, and whether each was an acquisition or a disposition.

Options and technicals tools

Options chain: a recent options-chain snapshot with prices, greeks, implied volatility, bid and ask, and open interest.
Options metrics: derived options figures such as implied-volatility percentiles and weighted call and put values.
Technical indicators: the full technical dashboard, grouped by panel (moving averages, MACD, ADX, Aroon and Parabolic SAR, Bollinger Bands, RSI, momentum, volatility, volume flow, pivot points, and advanced oscillators).

Portfolio tools

My portfolios: lists your own portfolios plus the GNG model portfolios you can view.
Portfolio holdings: the positions inside a portfolio with cost basis, gain or loss, allocation, and sector, for both manual and brokerage-connected portfolios.
Portfolio history: the transaction history (buys, sells, dividends, splits).
Portfolio performance: returns, a cumulative performance index, and cash against market value.
ETF look-through: your true exposure to a ticker across your portfolios once each held ETF is unpacked into its constituents, a single level of unpacking (an ETF held inside another ETF is not opened again). Focus one ticker ("how much NVDA do I really own, counting my ETFs") for a direct plus indirect total, or ask broadly for a table of your top look-through exposures with a per-ETF coverage readout. It works across every portfolio you own or collaborate on, manual and brokerage-connected, or scoped to one portfolio. When a fund's holdings are not covered by the platform's data, the answer names it as uncovered and never counts it as zero.
Watchlist management: reads and updates your watchlists on request, listing them or adding and removing a ticker.

Backtesting and simulation tools

Historical backtest: tests a portfolio or a basket of tickers across a past period, returning total return, compound annual growth rate, the Sharpe ratio, maximum drawdown, the final value, and a benchmark comparison.
Monte Carlo simulation: forecasts a range of future outcomes across thousands of simulated paths for a portfolio or a basket.

Research and content tools

Stock screener: runs a multi-filter screen across the research universe and returns the matching names. See Screens and how their data is produced.
Articles by stock: published GNG research articles tagged with a ticker, cited inline with a clickable link.
Articles by author: published GNG research by a named analyst.
SEC filings search: searches the text of public company filings from EDGAR and returns cited passages.
Documentation search: searches this documentation site and cites the pages it draws on, for how-do-I questions about the platform.

Memory tools

Read memory file: opens one of your saved memory documents by name.
Search memory and past chats: finds information across your memory files and your prior chat summaries.
Read past chat: fetches a previous chat by id or title.
Unified recall: a single search that fans out across your memory files, attachment contents, saved notes, and past-chat summaries when the source is unknown.
Remember: saves a fact you tell it to keep, so future chats retain it.

Visuals and creative tools

Chart and table cards: drops any inline widget card into the reply. See Widget types.
Generate image: creates an image from a description. Images can also be created in the background, arriving as their own later message while you keep chatting, or created directly for use inside a document rather than shown in chat. Images made for a document still appear in your Library.
Generate document: builds a downloadable PDF, PowerPoint, Word, or Excel file and returns a document card with a download button. See Generated documents.

Subagents

Delegate task: hands a sub-task to one of your specialist subagents. This option appears only when you have at least one subagent set up. The assistant can hand several sub-tasks off at once, and the parallel specialist runs return together, each reporting its own cost in the credits line. A per-answer delegation budget applies.
Manage agents: on request, the assistant lists, inspects, creates, updates, enables, or disables your own subagents. Creating an agent, or changing its instructions, description, tool set, or memory access, always shows you a preview and waits for your confirmation in chat before it saves. Deleting an agent is deliberately not available in chat; use the manager instead.

Skills

Skills: when skills are on for your account, the assistant can list your skills, apply one, or draft and update one for you, with confirmation before any change. See Skills.

Inputs

You do not fill in tool fields. Your only input is the plain-English request you type or speak in the message box. The assistant decides which tool to run and supplies the parameters from your request and the chat context. A few behaviors affect which tools are available.

The reference sheet is a read-only catalog, not a control panel.
The delegate option appears only when you have at least one subagent in your catalog.
Some tools are availability-gated for your account; when a tool is off, the assistant does not use it.

Outputs and how to read them

Tool results come back two ways, described in the following list.

Inline widget cards: charts, tables, gauges, and summaries shown at the point in the reply where they are relevant, each paired with a short written interpretation. For how to read each card, see Widget types.
Text in the reply: a cited article link, a recalled fact, or a confirmation that something was remembered.

While a turn is working, the console can show which tools ran. Memory-recall lookups can collapse into a single recall line, and other tool calls can show a brief running or done indicator. When an action needs a higher tier than you hold, the result is replaced by an inline upgrade card that names the gated feature and links to billing, rather than a raw error.

How it is calculated

At the tools level there is no single formula. The tools are data lookups and analyses, and each analysis is documented with its matching tool or widget. Where a tool runs standard, transparent math on your own data, that math is shown in full.

The backtest tool computes total return and the compound annual growth rate (CAGR) on your own holdings using standard definitions. Total return is the end value divided by the start value, minus one. CAGR annualizes that growth over the number of years.

CAGR = ( end value / start value ) ^ ( 1 / years ) − 1

In this formula, end value is the portfolio value at the close of the period in dollars, start value is the value at the start in dollars, and years is the length of the test in years. For example, a basket that grows from 100,000 to 180,000 over five years has a total return of 180,000 / 100,000 − 1 = 80 percent, and a CAGR of (180,000 / 100,000) ^ (1 / 5) − 1, about 12.5 percent per year. The screening and ratings tools surface GNG's own scores (the GNG Rating, GNG Quant, Safety, Quality, and Fair Value). Those scores are reported by the tools, but their internal formulas, weights, and thresholds are proprietary and are not shown.

Data and timing

The tools read the platform's stored data, built from public sources. Pricing, quotes, news and sentiment, options, earnings, and technicals trace to market data via AlphaVantage (Nasdaq quotes delivered through AlphaVantage). Company filings and fundamentals trace to EDGAR. Portfolio tools read your own portfolio data: manual portfolios update as you edit them, and brokerage-connected portfolios refresh on their regular sync. Memory and recall tools read only your own saved notes, attachments, and past chats. The assistant uses the most recent stored values at the time you ask; it does not stream live tick data.

Use cases

Ask whether a company's dividend is well covered, and let the console pull the profile, the dividend history, and the relevant scores, then chart them.
Ask the console to backtest one of your portfolios against the S&P 500 over five years and get the metrics and an equity curve without opening the Backtester.
Ask for dividend stocks above a target yield with analyst upside and get a screen returned as a sortable table.
Ask the console to recall what a prior chat concluded about a ticker, and let unified recall search your memory and past chats.

Limitations and disclosures

The tools surface valuations, ratings, forecasts, screens, and portfolio results. Treat their output as research, not a recommendation. The assistant works from stored data current to the platform's normal refresh schedule, not a live tick feed, and a heavier action can enforce its own per-tier gate at run time.

GNG Research provides equity research and educational tools, not investment advice. Nothing on the platform is a recommendation to buy or sell any security. Do your own research and consider your circumstances before making any investment decision.