> ## Documentation Index
> Fetch the complete documentation index at: https://takeprofit.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Chat Page

> A full-screen workspace for the AI agent — it analyzes tickers, writes Indie scripts, manages alerts, and hands you back charts you can keep working with.

The **AI Chat Page** at [takeprofit.com/ai](https://takeprofit.com/ai) gives the assistant a screen of its own. The conversation sits on the left, and everything the agent produces — charts, scripts, alert lists, backtests — opens on the right, where you can work with it directly.

<video
  autoPlay
  muted
  loop
  playsInline
  className="w-full rounded-2xl"
  style={{
border: "1.9px solid rgba(179, 200, 224, 0.08)",
objectFit: "fill",
display: "block"
}}
  src="https://mintcdn.com/takeprofit-a5dc0462/tR6NCsLYS0LS1oLO/images/guide/platform/AI/9.9_AI_Chat.mp4?fit=max&auto=format&n=tR6NCsLYS0LS1oLO&q=85&s=84c6f701c4023fa8da84cefafc39c822"
  data-path="images/guide/platform/AI/9.9_AI_Chat.mp4"
/>

This is a different surface from the [assistant inside your workspace](/docs/guide/platform/ai-assistant/AI-in-workspace): there you attach a widget and ask about it, here the agent builds what it needs as you talk. Both draw on the same usage budget.

## Sending your first request

Type your request in the composer at the bottom of the screen. Two controls sit under the field:

* **Skill picker** (left) — leave it on **Auto** to let the assistant choose, or open it and tick the skills you want it to use. See [AI tools](/docs/guide/platform/ai-assistant/How-to-work-with-AI-tools) for what each skill covers.
* **Model picker** (right) — which model answers you.

If you are not sure what to ask, use the suggestion chips under the composer. Pick a category — **Indicators**, **Technical analysis**, **Alerts**, or **Platform docs** — and it expands into ready-made starting points such as *Build an indicator*, *Pine to Indie*, *Analyze a ticker*, *Support & resistance*, *Create in bulk*, or *Set up a webhook*. Selecting one fills the composer, so you can edit it before sending.

You can paste attachments straight into the field. While an answer is streaming, keep typing if you want — messages are queued and sent once the current one finishes. **Stop** ends a response early and keeps whatever has already been written.

While the agent works, a compact progress line shows what it is doing and how many steps it has taken; expand it to see the individual steps.

## Threads

Every conversation is a thread, and threads persist — you can leave and come back at any time.

* **Recent threads** are listed in the sidebar, grouped by how long ago you used them.
* **Pin** the ones you keep returning to so they stay at the top.
* **Rename** or **delete** a thread from the same list.
* Each thread has its own address, so you can bookmark it. Browser back and forward move between threads.

The sidebar and the right-hand panel can both be resized, and the widths are remembered.

## Working with the answer

Answers are interactive. Rather than plain text, the agent replies with blocks you can dig into.

### Ask about a detail

Select any part of a reply — a term, a number, a summary card — and an **Explain with AI** action appears. It asks a follow-up about exactly that fragment, so you do not have to restate the context.

### Follow-up suggestions

After many answers the assistant offers next steps as chips — for example *Inspect TLT alerts* or *Reactivate BTC alert* after it lists your alerts. One click continues the conversation along that path.

### Charts

When your question is about price, the answer comes back as a chart rather than a description of one.

* A chart appears as a card in the conversation, with the indicator or drawings the agent added already applied.
* **Expand** it for a full-size view.
* **Edit** to change the symbol, timeframe, or drawings yourself — the agent keeps working with the chart as you leave it.
* **Copy to workspace** to move the chart into a workspace of your own and carry on there with the rest of your tools.
* Charts can also open as **tabs** in the right-hand panel, where the agent updates or closes them as the conversation moves on. The tab layout is saved per thread, so reopening an old conversation restores the charts that belonged to it.

### Structured results

Depending on the question, the agent renders purpose-built blocks instead of prose — a multi-timeframe structure summary with a card per timeframe, statistics and price readouts, tables, gauges, risk-reward and position-level views, scenario breakdowns. When it needs one more detail before continuing, it asks with a small inline form rather than a paragraph of questions.

Strategy questions can come back as a **backtest card**, which expands into a full backtesting tab.

## Indie scripts

Code the assistant writes is not just text to copy:

* **Open in IDE** turns a code block into a real editor tab with the Indie IDE inside the chat.
* The agent can write directly into that tab, streaming the script as it goes — useful for converting an existing script or fixing one that does not compile.
* **Add to chart** compiles the script into an indicator and plots it, either inline or in a chart tab.
* Several IDE tabs can stay open at once.

For the IDE inside a workspace, see [AI Assistant in the IDE](/docs/guide/platform/ai-assistant/AI-assistant-IDE).

## Alerts

You can ask the assistant to list, create, edit, pause, and delete [alerts](/docs/guide/alerts/Alerts-overview) without leaving the conversation. Listing them returns a table with the instrument, condition, delivery channels, status, and creation date for each one, so you can spot duplicates or alerts that never fired.

Bulk requests are supported too — applying the same condition across a list of symbols, or editing a group of existing alerts at once.

## Chat settings

Open **Chat settings** to shape how the assistant answers you. These are preferences, not hard rules — the assistant treats them as guidance.

**Responses**

| Setting             | What it does                                                                                                                      |
| ------------------- | --------------------------------------------------------------------------------------------------------------------------------- |
| **Explain terms**   | **On first use** explains trading terminology the first time it appears; **I know them** skips the explanations.                  |
| **Answer length**   | How much detail you want in a reply.                                                                                              |
| **If unclear**      | Whether the assistant asks you a clarifying question or makes its best assumption and continues.                                  |
| **Follow-up ideas** | Show or hide the suggestion chips after an answer.                                                                                |
| **Trading horizon** | The timeframe your questions are usually about — **Any**, **Intraday**, **Swing**, or **Position**. Analysis is framed around it. |

**System models** — the models available to answer, currently Claude Sonnet and Claude Haiku, with the active one marked. The model picker in the composer switches between them per message.

**API key** — connect your own Anthropic API key. Once enabled, every Claude model runs through your account instead of the hosted budget, which removes the TakeProfit usage caps. See [limits](/docs/guide/platform/ai-assistant/AI-assistant-overview#usage-and-limits).

**Delete all chats** removes your entire chat history. It cannot be undone.

## Next step

<Card title="How to Work with AI Tools" icon="wrench" href="/docs/guide/platform/ai-assistant/How-to-work-with-AI-tools">
  Technical Analysis, Indie Scripts, Alerts, Platform Guide, and Risk & Position Sizing — what each skill is for and how to prompt it.
</Card>