# The side panel

Use the message box, attach files, share tabs, talk instead of typing, save prompts as shortcuts and set your language.

The side panel is where you talk to the agent. Each tab has its own panel. To open it, click the Nanobrowser icon in the toolbar or press `Alt+Shift+N`.

## What's in the message box?

Type your task in the box that says **Ask anything…** and press `Enter`. Under the box are these controls:

- **`+`**: attach files or add a **Screenshot of this tab**. See [Attach files](https://nanobrowser.ai/docs/side-panel#attachments).
- **Mode**: **Ask** (**Ask before changes**) or **Auto** (**Make changes on its own**). See [Ask and Auto mode](https://nanobrowser.ai/docs/permissions).
- **Model**: the model for this chat, with a **Reasoning** row for its effort. See [Choosing a model](https://nanobrowser.ai/docs/models).
- **Microphone**: talk instead of typing. See [Use voice input](https://nanobrowser.ai/docs/side-panel#voice).
- **Send**, which turns into **Stop** while the agent works.

Each chat keeps its own mode and model. The ones you pick last also become the defaults for new chats. Your other chats keep theirs. You can also set the defaults in **Settings**.

## Share another tab with `@`

The agent works in the tab you started in and in the tabs it opens. To give it another tab:

1. Type `@` in the message box.
2. Pick a tab from the list. Type after the `@` to filter it.

The tab now belongs to this chat. It shows above the message box and joins the chat's tab group. To take it out, click the × on its chip. The tab stays open.

The list only shows web pages in the current window. A tab on a site you blocked is not shared. If another chat uses the tab, the panel asks before it moves it. See [Move a tab to another chat](https://nanobrowser.ai/docs/tasks#move-a-tab).

### Quote text from a page

Select text on a page, right-click it and choose **Ask Nanobrowser about this**. The text shows as a quote above the message box and goes with your next message.

## Attach files

You can add files in three ways:

- Click **`+`**, then **Attach files**.
- Drag files onto the message box.
- Paste an image or a file into the message box.

What you can attach:

- **Images:** PNG, JPEG, WebP and GIF. The model must see images. If it can't, the file stays with the chat but isn't sent.
- **Text files:** plain text, Markdown, CSV, JSON, HTML, XML, YAML, logs and code files.

PDF and Office files (Word, Excel, PowerPoint) can't be read in this version. Copy the text into a text file, or paste it into the message box.

Limits:

- Each file can be up to 25 MB.
- One chat can hold up to 100 MB of files in total.

You can't attach files while the agent works. Wait until it stops, or click **Stop**.

**Screenshot of this tab** in the `+` menu adds a picture of the current tab. It doesn't work on a tab where the agent is working, or on the browser's own pages.

The agent can also upload your attached files to a page, for example to a form. An upload counts as a change, so in Ask mode it asks first. See [Ask and Auto mode](https://nanobrowser.ai/docs/permissions).

## Save results to a file

### Ask the agent to save a file

Ask for a file, for example "save the results as a CSV file". The agent can save CSV, Markdown, JSON, plain text and HTML files of up to 5 MB to your Downloads folder.

- In Ask mode, a **Save this file?** card shows a preview. Click **Save**, **Save as…** to pick a folder and name, or **Skip**.
- In Auto mode, the file is saved without asking.
- The first time, Chrome asks you to let Nanobrowser save files. You can also allow **Saving files** in **Settings → Permissions**. Until you do, the card asks in Auto mode too.
- A file with the same name is never overwritten. A number is added to the new one.

When the file is saved, the chat shows **Saved** and the file name, with **Open** and **Show in folder**.

### Save a table from a reply

Under every table in a reply there are two buttons:

- **Copy** copies the table, ready to paste into a spreadsheet.
- **Save as CSV** downloads it as `table.csv`. This doesn't use the agent, and it works in any mode.

## Use voice input

You can speak your task instead of typing it. Voice input uses a voice model from OpenAI or Google Gemini, with your own API key.

To set it up:

1. Add an OpenAI or Google Gemini provider with its API key in **Settings → Providers**.
2. In the **Voice input** card on the same page, choose a **Voice model**: **GPT Transcribe** or **Gemini 3.5 Transcribe**.
3. Optional: set **Language** to the language you speak. It's **Detect automatically** by default.

To use it:

1. Click the microphone in the message box. The first time, a page opens to ask for the microphone. Click **Allow the microphone**.
2. Speak. Click the microphone again to stop. A recording stops by itself after 2 minutes.
3. Your words show up as text in the message box. Check them, then send.

If no voice model is chosen, the panel says **No voice model** and offers **Open Settings**. If the microphone is blocked, allow it in **Settings → Permissions → Microphone**.

## Save prompts as shortcuts

A shortcut is a prompt you save and start by typing `/`.

To make one:

1. Go to **Settings → Shortcuts** and click **New shortcut**.
2. Fill in:
   - **Name**: what you type after `/`. Lowercase letters, numbers and hyphens, up to 64 characters.
   - **Description**: optional. It shows in the list.
   - **Prompt**: the message the shortcut sends.
   - **Start page (optional)**: an address the agent opens first, in the current tab.
   - **Mode** and **Model**: optional. Leave them on **Keep the conversation's** to use the chat's own.
3. Click **Save**.

To use one, type `/` in the message box. Pick a shortcut with the arrow keys and press `Enter` or `Tab`, or click it. Type more letters to filter the list, or press `Esc` to close it.

In **Settings → Shortcuts**, click the star next to a shortcut to pin it. Pinned shortcuts come first in the list, and the first three show in an empty chat in place of the starter ideas.

### `/new` and `/compact`

The `/` list also has two commands:

- `/new`: start a new chat.
- `/compact`: sum up the older steps of this chat now, so a long task can go on. See [Long tasks](https://nanobrowser.ai/docs/tasks#compaction).

## Tell the agent about yourself

**Custom instructions** are what the agent should know in every chat. For example: "My company is Acme GmbH. Answer briefly."

Write them in **Settings → Memory → Custom instructions**. They can be up to 1500 characters. They save as you type.

Only you can change them. The agent can't.

## Change the language and theme

These are in **Settings → General**.

- **UI language** is the language of Nanobrowser's own buttons and menus: English, Simplified Chinese (简体中文) or Traditional Chinese (繁體中文). **System default** follows your browser.
- **Answer language** is the language the agent replies in. It's **Same as my message** by default. You can also pick English, Simplified Chinese, Traditional Chinese, Japanese, Korean, Spanish, French or German.
- **Theme** is **System default**, **Light** or **Dark**.

## Which keys can I use?

On a Mac, use `Cmd` where the table says `Ctrl`.

| Keys | What it does |
| --- | --- |
| `Alt+Shift+N` | Open the side panel (`Option+Shift+N` on a Mac) |
| `Enter` | Send. While the agent works, the message waits until its current step ends |
| `Shift+Enter` | New line |
| `↑` | In an empty message box, bring back an earlier message |
| `Ctrl+/` | Go to the message box |
| `Ctrl+Shift+P` | Switch between Ask and Auto mode |
| `Esc` | Stop the agent |
| `/` | Open the list of shortcuts and commands |
| `@` | Share a tab |

When an approval card has focus, `Enter`, `Ctrl+Enter` and `Esc` answer the card instead. See [Approval cards](https://nanobrowser.ai/docs/permissions#approval-cards).

To change `Alt+Shift+N`, go to **Settings → General → Keyboard shortcut** and click **Configure**.