> For the complete documentation index, see [llms.txt](https://help.intangible.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.intangible.ai/chat-agents-intangible/chat-with-intangible.md).

# Chat with Intangible

Open the chat panel, write a request to Intangible agents, name the card you mean, attach images, and answer the questions they ask with one click.

Tell Intangible agents what you want in plain words, for example "Render shots 02, 03 and 05 as video, each from its keyframe and scene." Type your request in the field at the bottom of the chat panel and send it, or click the microphone (**Start voice input**) and say it. Your browser asks for microphone access the first time. The field reads "Describe a scene, watch AI build it in 3D" until you type.

The chat panel sits in the bottom-right corner of the canvas and of Build. It is open when you land on a canvas, including a new project with no messages yet. The × collapses it to a bar, and it stays collapsed until you click the bar or the agents ask you a new question. Reloading the page opens it again. To see more of a long conversation, click the chevrons button beside the × (**Expand chat**): the panel grows to the full height of the window, older messages appear above and the message field stays at the bottom. Click it again (**Restore chat size**) to go back to the default height. The panel keeps its height when you collapse and reopen it, and goes back to the default when you reload. The button is hidden during a guided tour and on windows too short to grow into. When the agents are idle, the bar reads **Chat** over the chat's name, or **New chat** until you rename it. Otherwise the bar says what they are doing: **Working on your request** while they work, **Waiting for your answer** when they have a question, and **Ready** when they finish, until you open the panel. While they reply, a light runs around the panel's edge. Compose and Visualize have no chat panel; ask from the canvas or Build for work in those modes, such as a camera move or a render.

## Point at what you mean

Name the thing you mean the way the canvas names it: "shot 02's keyframe", "scene 01", "the crater reference". The agents read the canvas, so a precise name is enough for them to find the card. A card's name is the title on the card: the name it was given, otherwise the file name of an upload, or for a render, its scene and shot names with a version number. Selecting a card does not point the agents at it, so name it.

## Add images

Click **+** to attach images to a message, up to 10 at a time; you can also drop or paste them onto the field. You can send images with no text. Each message that sent images shows a line in the thread, such as **2 images**, under your text or on its own. When you send, the field waits while they upload, then they land in **Sources** and the agents look at them. Images can be jpeg, png or webp, each under 20 MB. If some fail to upload, the message still goes out with the ones that made it.

A card on the canvas cannot be attached to a message: **+** takes images only, so name the card instead.

### Guide me or Skip

The menu next to **+** holds the same choice as Home: **Guide me** or **Skip**. It is shared with Home, and you can switch it at any point in a project; the agents follow whichever is set when you send. With **Guide me**, the agents ask their questions, wait for you to confirm the brief, and get your approval of the scenes and shots before rendering anything. With **Skip**, they ask nothing: they make the open calls themselves and say in the brief which ones they made, then make the reference sheets, build the scenes and shots, and render an image of every shot in the same reply, which spends credits. Skip does not cover video. The agents make one only when you ask, and they still quote the cost and wait for you to agree. See [The brief](https://help.intangible.ai/canvas/the-brief.md).

## Answer their questions

When the agents need a decision, they ask with a **choice card**: a short question with a list of options. Click one to answer: the options go, and the question stays in the thread with your pick under it, so scrolling back reads question, answer, question, answer. When a question takes more than one answer, or several questions arrive together, pick your answers, then click **Send** under the list. Answering a choice card does not cost a new message. With Guide me on, the questions come one at a time, and the last one asks whether to confirm the brief or edit it first. With Skip on, there is no brief to confirm. See [The brief](https://help.intangible.ai/canvas/the-brief.md).

## Watch them work

While they work, the agents show a status line and a list of steps, each marked as queued, running or done. See [Status and steps](https://help.intangible.ai/chat-agents-intangible/status-and-steps.md).

On the canvas, the agents keep their replies short. When a turn builds a scene, places cameras or starts renders, they end it with two or three things they can do next, in example phrases you can reuse, and a short question.

## Rate a reply

Each answer has thumbs up and thumbs down under its last reply. Click the other thumb to change your rating, or the same one again to clear it. Use them: the ratings tell the team which answers worked.

## Where the conversation lives

Each person has their own conversation on a project; a teammate on the same project sees their own, not yours. It follows you across scenes and between the canvas and Build, and it is still there after you reload the page or come back to the project later. Chats from before the canvas, which were kept one per scene, are still saved but don't show in the chat panel. To name it, open the chat panel, hover the title, double-click and type a name of up to 120 characters; Enter or clicking away saves, and Esc cancels. The name stays after a reload and shows on the collapsed bar. Once you rename the chat, its name replaces the status line at the top of the open panel. The agents know which mode and scene you were in when you sent each message.

## Related

* [Chat](https://help.intangible.ai/chat-agents-intangible/chat.md)
* [Status and steps](https://help.intangible.ai/chat-agents-intangible/status-and-steps.md)
* [Pick a model](https://help.intangible.ai/chat-agents-intangible/pick-a-model.md)
* [Brief AI like a DP](https://help.intangible.ai/learn-more/how-to/brief-ai-like-a-dp.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://help.intangible.ai/chat-agents-intangible/chat-with-intangible.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
