# Tools and files in the chat

On its own, a model only writes text. Tools give it actions: find fresh information, open a page, read your document, put together a file. The model decides by itself when a tool is needed — you only need to ask in plain words.

## The tools menu

Click "+" in the message field and open "Tools". Each tool has a switch and a hint with a question mark.

| Tool | What it does | In a new conversation |
|---|---|---|
| "Current time" | Tells the model the current date and time | On |
| "Web search" | Searches the web and returns up to 10 results with links | On |
| "Read web pages" | Opens a page by its link and reads its text. An address that did not appear in the conversation is opened only with your consent | On |
| "Knowledge base search" | Looks for an answer in the project's [knowledge bases](https://support.watbot.org/knowledge-bases) | On |
| "File creation" | Creates PDF and Word documents, Excel spreadsheets, CSV and text files | On |
| "Extended formatting" | Allows colored callout blocks and marker highlighting in responses | Off |

![The “+” → Tools menu](https://support.watbot.org/_media/en/08-ai/_images/tools.webp?v=1791781230)

The set of tools is remembered separately for each conversation. The model cannot call a tool that is turned off.

Chatbot, mini app and mini landing page building are not in the menu: the AI connects them by itself when you ask to build or fix a bot, an app or a landing page, and a "Chatbot builder", "Mini app builder" or "Mini-landing builder" chip appears above the message field. In the same way, without a menu item, the AI connects ["Knowledge base changes"](https://support.watbot.org/ai/tools#knowledge-base-changes) when you ask to save, add or delete something in a knowledge base. Of all the tools, only building and knowledge base changes change project data. How it turns on and how the "Ask first" and "Automatic" modes differ is described on the [Building a bot in the AI chat](https://support.watbot.org/chatbots/ai-builder) page; mini apps — on the [Building an app in the AI chat](https://support.watbot.org/mini-apps/ai-builder) page, mini landing pages — on the [Building a landing page in the AI chat](https://support.watbot.org/landings/ai-builder) page.

Widget building connects by itself in the same way, with a "Widget builder" chip. See the [Building a widget in the AI chat](https://support.watbot.org/widgets/ai-builder) page.

Booking setup connects by itself in the same way, with a "Booking setup" chip. See the [Setting up booking in the AI chat](https://support.watbot.org/booking/ai-builder) page.

Online school building connects by itself in the same way, with a "School builder" chip. See the [Building a course in the AI chat](https://support.watbot.org/schools/ai-builder) page.

> [!NOTE]
> Each web search is charged to "AI Credits" separately, on top of the model's tokens. The other tools have no separate price, but their results go into the request to the model and increase token usage.

### Reading pages at unfamiliar addresses

The model opens links from the conversation by itself: addresses from your messages, from web search results, from pages it has read, and from what it found in the knowledge base and the help center. An address the model built itself is not opened without you. The conversation shows an "Open a page" card with the address and the note "this address did not appear in the conversation — the model built it itself. Open it only if you trust the link". The page is loaded only when you click "Open"; "Reject" cancels the loading.

The card appears every time, including in the "Automatic" mode, and it has no "Apply and don’t ask again" button. This protects the chat from hidden instructions in someone else's text: a page or a document can tell the model to open a third-party site's address with your data added to it. The rule works only in the AI chat — the "AI Agent" block has no such card.

### Help center search

For questions about the platform itself — how to set up a block, connect a channel, what a service costs — the model looks for the answer in this help center and gives links to its sections. This tool is not in the menu: it is always on and has no separate price. In the response, the call shows up as a "Searching the help center" step with the list of sections found.

### What it looks like in a response

Tool calls are shown in the response as a group of steps: "Searching the web", "Reading the page", "Searching the knowledge base", "Creating a file". You can expand the group to see exactly what the model searched for and what it got: the sites found, knowledge base fragments with the document name, the lines of the file it read.

## Attachments

You can add a file to a message in four ways:

- "+" → "Upload file";
- drag the file into the chat window;
- paste it from the clipboard — for example, a screenshot;
- "+" → "From library" — choose from files already uploaded to chats in this project.

### Which files AI can read

- **Documents:** PDF, Word (DOC, DOCX), PowerPoint (PPT, PPTX), Excel (XLS, XLSX), OpenDocument (ODT, ODP, ODS), RTF, EPUB.
- **Text files** with any extension: TXT, CSV, JSON, code, SVG and others — anything whose content is text.
- **Images:** JPG, PNG, WebP, BMP, TIFF, GIF, HEIC — only if the selected model recognizes images.

The model reads a document in parts when it needs them, so a large file does not go into the request in full. While a document is being prepared for reading, the chat shows "Processing documents…".

A document can be sent without text. An image needs a question added — otherwise sending is unavailable: "Add text to the image".

Clicking a document in a message opens it to the right of the conversation: the text with a "View" / "Source code" switch and the "Download" and "Copy" buttons.

### Attachment limits

| What | Limit |
|---|---|
| Files per message | 10 |
| Files per conversation | 1,000 |
| Image size | 5 MB |
| Document size | 50 MB |

The chat will not accept a document larger than 50 MB and will suggest uploading it to a [knowledge base](https://support.watbot.org/knowledge-bases): there, only the relevant fragments go into the request to the model, not the whole text.

### Images and the model

If the selected model does not recognize images, you will not be able to attach a picture: "The … model does not support images". In the "Project files" window, such files are marked "a vision model is required".

If the conversation already has pictures and you switch to a model without image recognition, a warning appears under the input field: "The selected model does not recognize images — pictures from this chat are unavailable to it".

## Creating files

Ask the model to make a file — "make an estimate in Excel", "format this as a contract in Word", "export the list to CSV". The finished file appears in the response as a card with the name, format and size.

| Format | What you get |
|---|---|
| PDF | A document with headings, lists and tables — for a finished, final version |
| DOCX | A Word document with the same formatting — if it will be edited |
| XLSX | An Excel workbook: sheets, a table header, formulas |
| TXT, MD, CSV, TSV, JSON, XML, HTML, YAML, ICS, VCF, SQL | A text file with the content as is |

Clicking the card opens the file content to the right of the conversation. The "Download" button on the card saves the file to your computer.

### Editing a created file

You can ask the model to change a file it created earlier in this conversation: "change the price in the third row", "add a section about delivery". The model makes the edits and returns the new version as a separate card.

Only files created by the model itself can be edited this way. The model will read a document you uploaded but cannot change it — ask it to create a new file based on it.

### Limits

- No more than 5 files per response.
- Images are not inserted into PDF and Word documents.
- An Excel spreadsheet — up to 20 sheets and up to 50,000 cells.
- The model does not run programs or convert files: the whole file is made from the text it wrote.

## Knowledge base changes

Ask to save, add or delete something in a knowledge base, and the AI connects knowledge base changes by itself, like building, and a "Knowledge base changes" chip appears above the message field. The model adds a new document: in the "Ask first" mode after the "Add to the knowledge base" card, in the "Automatic" mode at once. It deletes only after you confirm on a card, in either mode. When the changes are connected, what they need and what they cannot do is on the [Connect a base to AI](https://support.watbot.org/knowledge-bases/connect-ai#knowledge-base-changes) page.

## If something does not work

- **The model answers in text, but there is no file card.** Check that "File creation" is turned on in the "Tools" menu. A file has been created only if the response has its card.
- **"File is too large for the chat".** The document is larger than 50 MB. Upload it to the project's knowledge base.
- **"The image "…" is larger than 5 MB. Reduce its size and try again."** Compress the picture and attach it again.
- **"You can attach no more than 10 files at a time".** Send the files in several messages.
- **"You can send no more than 1000 files in a single chat. Create a new chat to continue."** Start a new conversation.
- **The model does not search the knowledge base.** Check that "Knowledge base search" is turned on and that the base has documents in the "Ready" state.
- **An "Open a page" card, although you gave the link yourself.** Most likely the model changed the address, for example added parameters to it. Compare the address on the card with yours and click "Open" only if you trust it.
- **The model does not change the knowledge base.** What to check is on the [Connect a base to AI](https://support.watbot.org/knowledge-bases/connect-ai#if-something-does-not-work) page.
