> ## Documentation Index
> Fetch the complete documentation index at: https://basedash.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Chat
> Ask questions about your data in plain English and get answers with the SQL behind them
Chat is Basedash's AI data analyst. You ask a question in plain English, and the AI writes SQL, runs it against your connected data, and answers with a chart, a table, or a short write-up.
> [!NOTE]
>
> Available on all plans. Admins and members can chat. [View-only
> members](https://basedash.com/docs/administration/members-and-roles.md#view-only) can read chats shared
> with them but can't ask questions.
## Overview
Use chat for questions you haven't answered before, or when you want to explore a topic before building a dashboard. Each answer shows the queries the AI ran, so you can check how it got there. When an answer is worth keeping, add its chart to a [dashboard](https://basedash.com/docs/features/dashboards.md). For questions you want answered on a schedule, use [automations](https://basedash.com/docs/features/automations.md).
Chat uses everything Basedash knows about your organization: the schema of each [data source](https://basedash.com/docs/data-sources/introduction.md) you can access, [models](https://basedash.com/docs/features/models.md), [AI context](https://basedash.com/docs/features/ai-context.md), and [skills](https://basedash.com/docs/features/skills.md). The better that context is, the better the answers get. See [Accuracy and trust](https://basedash.com/docs/features/accuracy-and-trust.md) for how answers are generated and checked.
You can also chat with Basedash from [Slack](https://basedash.com/docs/features/slack.md), from AI tools through the [MCP server](https://basedash.com/docs/features/mcp-server.md), and from your phone with [Basedash Mobile](https://basedash.com/docs/getting-started/basedash-everywhere.md#basedash-mobile).
## Start a chat
1. Click **Chats** in the sidebar, then click **+** to open a new chat.
2. Type your question in the input under **What questions do you have?** and press Enter.
3. Ask follow-up questions in the same chat. The AI keeps the earlier questions and answers in mind.
A new chat shows a few suggested questions based on your recent chats, dashboards, and data. They refresh every week. Click one to ask it, or click **Dismiss** to hide it.
Good questions name the metric, the time period, and any filters:
* "Show me weekly signups over the last 6 months, broken down by plan"
* "What are our top 10 customers by revenue this year?"
* "Which customers haven't placed an order in the last 30 days?"
* "Why did churn go up in September?"
## Read an answer
While the AI works, it shows what it's doing, such as the query it's running or the model it's reading. When it finishes, the answer shows how long the analysis took, for example "Analyzed for 24s".
Open **Sources** under an answer to see each query the AI ran. For each query you can see the data source, the number of rows returned, a preview of the first rows, and the full SQL with a copy button.
To stop an answer early, click **Stop generating**. Whatever the AI has written so far stays in the chat.
Rate an answer with **Good response** or **Bad response**. Feedback helps the Basedash team improve answer quality.
## Attach files
You can attach files to a question and ask the AI to analyze them alongside your data.
1. Click **+** in the chat input and choose **Upload files**. You can also drag files into the chat or paste them.
2. Ask your question about the files.
Supported files include images, PDFs, CSV and TSV, Excel, Word, PowerPoint, JSON, Markdown, and plain text. Each upload can be up to 50 MB, including all files attached at once.
Attached files are only available in that chat. To query a file from charts and dashboards, connect it as a data source instead. See [Can I upload my own CSVs of data?](https://basedash.com/docs/help/faq.md#can-i-upload-my-own-csvs-of-data).
## Ask with your voice
Click **Start voice recording** in the chat input, speak your question, then click **Finish voice recording**. Basedash transcribes the recording into the input so you can review it before you send it. Recordings stop after 5 minutes.
## Save charts and build dashboards
When an answer includes a chart, click **Add to dashboard** and choose an existing dashboard, **New dashboard**, or **Drafts**. The chart keeps its SQL, so it stays up to date on the dashboard. The chart's menu also lets you copy it as an image, copy its SQL, or export its data.
You can also ask chat to build a whole dashboard, for example "Build a dashboard of our sales pipeline by stage, owner, and month". The AI creates the dashboard with charts, tabs, and filters, and shows a live preview in the chat. Click **Open dashboard** to keep editing it.
## Download data as CSV
Ask for the data as a file, for example "Export all orders from last month as a CSV". The AI runs the query and attaches a CSV that you can download from the chat.
* CSV files can contain up to 1,000,000 rows. If the result is larger, the file is cut off and marked **Row limit reached**.
* The query runs with your data access, so the file only contains data you're allowed to see.
* Download links expire after 30 days.
## Approve actions
Some actions need your approval before the AI runs them. When that happens, the chat shows a card describing the action, with **Approve** and **Deny** buttons.
* **Tools from MCP connectors.** Admins choose which [MCP connector](https://basedash.com/docs/features/mcp-connectors.md) tools always run, need approval, or are blocked.
* **Changes to your database.** The AI can propose a single `INSERT`, `UPDATE`, or `DELETE` statement on Postgres, Redshift, and Supabase data sources where you have edit access. The card shows the SQL, and nothing runs until you click **Approve and run**.
Approvals only work in the web app. See [Read-only by default](https://basedash.com/docs/features/accuracy-and-trust.md#read-only-by-default) for details.
## Share a chat
Click **Share** in a chat to open **Share chat**. You can share with everyone in your organization, with [groups](https://basedash.com/docs/administration/groups.md), or with specific members.
| Level | What it allows |
| - | - |
| **Can view** | Read the existing messages |
| **Can edit** | Read the messages and continue the conversation |
| **Full access** | Everything above, plus manage access and delete it |
The creator of a chat always has full access, and organization admins always have full access to every chat.
### Default chat access
By default, new chats are shared with everyone in the organization with full access. Admins can change this in **Settings → Chat → Default access**:
* **Default audience**: **Everyone**, **Creator only**, or **Creator's groups**.
* **Default permission level**: **Full access**, **Can edit**, or **Can view**.
These defaults only apply to chats created after the change.
## Manage chats
Open a chat's menu to **Rename** it, **Favorite** it, **Copy link**, or **Delete** it. Deleting needs full access, and deleted chats can't be restored.
Chats started from Slack appear in your chat list with a **Slack** badge. Chats started through the [API](https://basedash.com/docs/api-reference/overview.md) or the MCP server don't appear in the list.
## Limits
> [!NOTE]
>
> The AI can take up to 50 steps to answer one question, such as running a query
> or reading a model. If it reaches the limit, it summarizes what it found and
> lists what's unfinished. In long chats, older messages are summarized so the
> conversation can continue. The most recent messages are always kept in full.
## FAQ
> [accordion] Can I see the SQL behind an answer?
>
> Yes. Open **Sources** under the answer to see every query the AI ran, with a
> preview of the results and a button to copy the SQL.
> [accordion] Who can see my chats?
>
> It depends on your organization's [default chat
> access](#default-chat-access). Unless an admin has changed it, new chats are
> shared with everyone in the organization. Use **Share** to change access for
> a specific chat.
> [accordion] Can I choose which AI model chat uses?
>
> No. Basedash picks and manages the models. Self-hosted deployments can
> configure their own AI provider. See [AI
> configuration](https://basedash.com/docs/self-hosting/configure.md#ai-configuration).
> [accordion] Why does chat say it can't find my data?
>
> The AI can only query data sources you have access to. Ask an admin to check
> your [data source access](https://basedash.com/docs/administration/data-source-access-control.md), and
> add [AI context](https://basedash.com/docs/features/ai-context.md) that explains where the data lives.
> [accordion] Does chat use AI credits?
>
> Yes. Chat counts toward your organization's AI usage. See [Billing and AI
> usage](https://basedash.com/docs/administration/billing-and-ai-usage.md).
## Related
* [Accuracy and trust](https://basedash.com/docs/features/accuracy-and-trust.md)
* [Dashboards](https://basedash.com/docs/features/dashboards.md)
* [AI context](https://basedash.com/docs/features/ai-context.md)
* [Slack](https://basedash.com/docs/features/slack.md)
* [Basedash everywhere](https://basedash.com/docs/getting-started/basedash-everywhere.md)
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.