Chat

The chat is the main workspace of the AI Console. You hold a threaded conversation, the assistant pulls live data and renders inline widgets, and you manage, share, branch, and print your chats. This page explains what the chat shows, the controls you set, how to read each output, and how an answer's credit cost is added up. It is available on the Free, Pro Lite, and Pro plans, with usage metered by GNG Credits.

Before you begin

Sign in to your account. The chat runs once you are signed in.
To send a file with a message, attach it directly in the composer. For the supported kinds and limits, see Attachments and file uploads. To reuse a document across chats, save it to your knowledge first: see Your knowledge and retrieval.

Quickstart

Open a chat
Go to the AI Console and start a new chat or select a saved one from the sidebar.
Ask a question
Enter a question in the message box and send it. The answer streams in, then attaches any relevant widget.
Read the credits row
Expand the credits row under the finished answer to see where the credits went.
Act on the answer
Use the per-message controls to copy, print, or regenerate the answer, or branch the conversation from a chosen message.

What it shows

The chat surface is built from a conversation pane, a composer, and a set of header and sidebar actions. The parts are described in the following sections.

Conversation pane

Your messages and the assistant's answers, in order. A long answer streams in as text.
Inline visual widgets render inside an answer: price charts, company overviews, dividend charts, earnings tables and charts, sentiment gauges, options chains, insider-transaction tables, comparison tables and charts, portfolio holdings and performance, screen-result tables, and backtest and Monte Carlo result panels.
A time stamp on every message. Your message is stamped when you send it, and an assistant reply is stamped when it finishes. On a phone the stamp shows the time only.
A single compact credits row under each finished answer, showing the total GNG Credits the answer used with a toggle to open it. Expanded, it lays out the cost breakdown and a numbered list, in run order, of the steps the answer ran.
A document card at the end of a reply that produced a downloadable file. See Generated documents.

Composer

A growing text box at the bottom with these controls:

Attach file (paperclip): attach images, audio, text, PDF, spreadsheet, and document files for the assistant to read. See Attachments and file uploads.
Mention a subagent (the @ button): opens a picker to route the next message to one of your subagents. See Subagents.
Record voice (microphone): records a short clip that is transcribed and attached to your message.
Send, which becomes Stop while a reply is streaming.

Slash commands

Type / at the start of a message to open the skill picker when skills are on for your account. See Skills. Type /help to see what the assistant can do, or /tools to list the live-data tools it can call. Both show an in-chat help card and do not spend a turn. Anything else you type is answered by the assistant.

Per-message actions

On the assistant's settled replies: Copy (copies the reply as plain text), Regenerate (re-runs the answer), and Print (renders that message for printing or saving as a document). A small credit-cost line shows what the turn cost.
On your own messages: an actions menu with Delete message and Print message, plus the ability to resend or branch the conversation from that point.

Header and sidebar actions

Model badge: shows the current model and opens Chat settings when selected.
Take the tour: a guided walkthrough of the console.
Share: available on a settled, non-empty regular chat. It creates a read-only link so any signed-in user can view that conversation. Incognito chats cannot be shared.
Print chat: renders the whole conversation for printing or saving as a document.
Sidebar: New chat, a searchable list of your chats, entries for Marketplace, Workflows, and, where available, War Room, a utility grid of Memory, Agents, Skills, Library, Settings, and Usage, and Back to GNG (returns you to the page you came from). The entry labeled Agents opens the subagents manager. On desktop, Cmd or Ctrl plus B collapses the sidebar.

Per-chat management

From each chat row's menu: Rename, Pin to the top, Share, an Incognito toggle, and Delete. Incognito deletes that chat's stored summary and stops new summaries for it, affecting only that chat.

Inputs

Message: free text, your question or instruction. Enter sends; Shift plus Enter adds a newline; Esc stops a streaming reply.
Attachments: staged files you attach to a message, such as images, audio, PDF, spreadsheets, and documents. See Attachments and file uploads.
Mentions: type @ to point the assistant at one of your subagents for a turn.
Main model (in Chat settings): the model that drives the conversation and decides when to call data tools. The default for a new chat is a mid-size model. Smaller variants typically cost fewer GNG Credits. Changing it in settings sets your default for every new chat and applies to the current chat.
Reasoning depth (in Chat settings): one of Low (fastest and cheapest), Medium (recommended default), High (deeper analysis), or Max (expensive, use sparingly). Default is Medium. Higher depth gives smarter answers at higher cost per message.
Privacy toggles (in Chat settings, all default on): Auto-capture memories, Auto-accumulate chat memory, Allow portfolio access, and Allow brokerage access (for an account you have securely connected). Turning one off blocks the assistant from writing or reading that surface. Existing data is never deleted by a toggle. See Memory.
Per-message actions: copy, print, and regenerate. Regenerate re-runs the last answer and consumes credits again.
Branch: fork the conversation from a chosen message into a new chat, keeping the earlier history.
Delete a turn: hides a message and its reply from the thread. The hidden turn is kept only for your usage history and is not shown again.
Share: produces a read-only link to a settled chat. Private and in-progress chats are excluded.

Outputs and how to read them

The streamed answer: reads top to bottom as it forms. A thinking indicator means the model is still working.
The steps list: expand the credits row under an answer to see a numbered list, in run order, of the steps the answer ran, so you can see what it was built from.
Inline widgets: charts, tables, and result cards appear inside the answer. A backtest or Monte Carlo answer shows the full result card in place.
The credits row: collapsed, it shows the total GNG Credits the answer used with a toggle to open it. Expanded, it breaks the cost into the main answer, web searches (priced per call), and each subagent that ran (with its own cost and run time), and lists the numbered steps the answer took in the order they ran.
Answers keep working when you leave: a long answer keeps running if you close the tab, switch chats, or lose the connection. Reopening the chat resyncs the progress so far, and the sidebar marks a chat whose answer finished while you were away.
Downloads never leave the chat: downloading a generated file or image opens in a new tab, so the conversation stays put.
Generated images enlarge: select a generated image to open a full-screen enlargement. Press Escape or select the backdrop to close it.
Background images: an image you ask for in the background arrives as its own later message in the chat, and that message carries its cost.

How it is calculated

An answer's credit cost is the sum of its parts. The pieces are described in the following list.

Main answer: the cost of the assistant composing the reply, including the data tools it calls. Tool calls other than web search are not billed as separate charges; the work they trigger is already counted inside the main answer's cost, so they appear by name with no separate figure.
Web searches: counted per call. Each web search the answer runs adds its own charge.
Subagents: each subagent that runs reports its own credit cost and run time.

The total under an answer is the main answer plus any web searches plus any subagents. For example, an answer whose main reply costs 40 credits, that runs two web searches at 5 credits each, and that delegates to one subagent costing 20 credits, totals 40 + 10 + 20 = 70 credits. The exact cost of each part depends on the model, the reasoning depth, and how many tools and subagents the answer uses, so the same question can total different amounts on different settings. For the user-facing meaning of the meter, see GNG credits.

Data and timing

The assistant pulls live company basics, prices and technical indicators, fundamentals, earnings, dividends, ownership, news and sentiment, and options data on demand, and it can read your own portfolios. Market and fundamentals data come from AlphaVantage (Nasdaq quotes delivered through AlphaVantage) and from EDGAR, current to the platform's normal refresh schedule. Your portfolio data reflects your latest saved or connected holdings.

Use cases

Hold a multi-step research thread on one company and keep the figures together.
Branch a conversation to explore an alternative without losing the original.
Print or share a finished analysis.
Point a turn at a specialist subagent with @ for a focused task.

Limitations and disclosures

Chat answers regularly present valuations, sentiment, screens, forecasts, and portfolio figures. Treat them as research, not a recommendation. Because credit cost varies with the model, the reasoning depth, and the tools used, the same question can cost different amounts on different settings.

The microphone in the composer records a short clip and transcribes it into your message, and it works for every plan. Real-time Voice mode is a separate capability and is not currently available to Free, Pro Lite, or Pro users. See Voice mode. The War Room is a multi-agent council inside the AI Console, available on the Pro plan and above and rolling out in waves, so an eligible account may not see it yet. See War Room.

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.