MindsDB Anton — Autonomous BI Agent

Anton is an AI agent built for getting things done. If you need to analyze a dataset, generate a report, or automate a workflow, Anton is for you.

Check out Anton GitHub repository here.

Onboarding

Getting started with Anton is straightforward — install it locally with a single command:

• On MacOS and Linux:

curl -sSf https://raw.githubusercontent.com/mindsdb/anton/main/install.sh | sh && export PATH="$HOME/.local/bin:$PATH"

• On Windows:

irm https://raw.githubusercontent.com/mindsdb/anton/main/install.ps1 | iex

Once installed, run Anton with the following command:

anton

Anton will then ask you to pick an LLM provider — think of this as choosing the AI engine that powers your experience. Here are the available options:

• Minds Cloud (Recommended) Minds Cloud is purpose-built to work seamlessly with Anton, automatically routing each task to the best model available, resulting in faster responses and optimized token cost. To get started, head over to mdb.ai, create an account, choose a plan, and generate your API key. Then just paste it into Anton when prompted.

• Bring Your Own Key (BYOK) Anton also supports third-party LLM providers such as Anthropic and OpenAI. To use this option, provide an API key from your preferred provider when prompted.

Here is how Anton greets you and walks you through the setup process:

Once you select the LLM Provider, Anton will prompt you for the API key.

/Applications/Python 3.12/Install\ Certificates.command

python3 -c "import urllib.request; urllib.request.urlopen('https://mdb.ai/')" && echo "SSL OK"

Connecting Data Sources

Anton runs locally on your machine and provides a secure bridge between your data sources and the underlying AI models. There are two ways to connect data:

• Secure connection (recommended) – Best for APIs with passwords, keys, or any sensitive credentials
• Direct connection – Best for public APIs with no sensitive credentials

Secure Connection

When you connect data securely, Anton collects your credentials and stores them as environment variables on your machine. This means the AI models do not see your sensitive credentials — they are kept entirely local to you.

Using a Predefined Template

Anton includes ready-made connection templates for popular data sources. The full list of available templates, along with their required and optional parameters, is in datasources.md.

Steps:

1. Run the following command:

/connect

2. Select a data source from the list.
3. Enter the required connection parameters when prompted.
4. Start asking Anton questions about your data.

Creating a Custom Template

If your data source isn't listed in datasources.md, you can describe it to Anton and it will build a custom connection template for you.

Steps:

1. Run the following command:

/connect

2. Select "Custom datasource (connect anything via API, SQL, or MCP)".
3. Describe your data source — what it is and how you want to connect to it. Anton will generate a template based on your description.
4. Enter the required connection parameters when prompted.
5. Start asking Anton questions about your data.

Direct Connection

You can also pass connection details directly to Anton in conversation. This is the simplest approach and works well when your credentials are not sensitive (for example, a public API that requires no authentication).

Managing Data Sources

Once a data source is connected, you can manage it using the following commands:

List all connected data sources

/list

Displays all data sources currently connected to Anton.

Edit a data source

/edit datasource_name

Update the connection parameters for an existing data source. Replace datasource_name with the name of the data source you want to edit.

Remove a data source

/remove datasource_name

Disconnects and removes a data source from Anton. Replace datasource_name with the name of the data source you want to remove.

Test a data source connection

/test datasource_name

Checks that Anton can successfully connect to a data source. Replace datasource_name with the name of the data source you want to test.

Quick Reference