> ## Documentation Index
> Fetch the complete documentation index at: https://ara-90a60a07.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Get started

> Start a task with Ara.

Read this guide at [docs.ara.so](https://docs.ara.so). [ara.so/docs](https://ara.so/docs)
opens the same documentation there.

1. Open [Ara](https://ara.so), review how the coding-agent loop works, and sign in with Google or an email verification code.
2. Choose **New task** and describe the result you want.
3. Name a repository or Device only when the task needs one. Otherwise, Ara can
   start in a scratch workspace.

For example:

```text theme={null}
Fix the flaky test in src/auth/session.test.ts.

Reproduce the failure, add a regression test, run the focused test, and open a pull request.
```

<img className="doc-screenshot" src="https://ara.so/demo/agent-session-light.webp" alt="An Ara task session showing the conversation and work panels." />

For recurring work, create an automation. For work that runs longer, return to
the session to follow progress, inspect evidence, and decide what happens next.
Add organization context when every task should begin with the same company
background.

## Choose an interface

| Need                                  | Use                                 |
| ------------------------------------- | ----------------------------------- |
| Start and steer a task                | [Ara web app](https://ara.so/login) |
| Work from your terminal or editor     | Ara CLI                             |
| Give an AI client access to Ara tools | Ara MCP                             |
| Work from a team conversation         | Slack with @Ara                     |
| Let a task use your Mac               | an approved Ara Device              |

To connect a Mac, install the CLI and follow the sign-in prompt:

```bash theme={null}
curl -fsSL https://ara.so/install.sh | bash
```

## Models and keys

Choose a model in the task composer, or connect a provider in **Settings →
Models & credentials**. Provider keys are encrypted and stay inside Ara’s
service boundary; they are not added to a task, terminal, or command line.

Ara uses the model you selected. If a provider is unavailable, the task shows
the failure instead of silently switching models. Model usage and sandbox time
draw from your workspace credit balance.

## Continue with your task

* Review the session to see the conversation, changes, commands, and evidence
  together before accepting a result.
* Use automations for repeated work on a schedule, webhook, or event.
* Connect repositories, plugins, and approved Devices only when the task needs
  that context.
* Keep credentials in the workspace settings rather than source code, prompts,
  or terminal history.

Ara carries the work from your request through a reviewed change, with the
conversation and evidence kept together in one place.