# Get started with Rukou

Connect one agent, give it a small browsing task, and watch the session in Rukou. You need a
Rukou invitation, access to a workspace, and an AI app that supports remote MCP servers.

## Open your workspace

Open the link in your invitation and complete the account setup. Sign in and choose a workspace.
If you cannot see the workspace you need, ask your organization owner for access. If your
organization is not on Rukou yet, [request access](https://rukou.ai/#access).

## Create and connect an agent

1. Open **Agents** in the workspace.
2. Enter a name under **New agent**, such as `Research`, and select **Create**.
3. In the setup that opens, choose your app under **Which app are you connecting?**
4. Follow the instructions shown for that app. Use the address and commands generated for
   this agent.
5. If your app uses sign-in, finish the Rukou authorization page. If it uses an API key, make
   a key for this agent and put it in your app's secure configuration.

The setup shows **This agent has an active connection** after an app signs in or uses its API
key. Creating a key alone leaves it waiting for the first connection.

To let your agent do the connecting, paste the [setup prompt](agent-setup.md#connect-this-app)
into your app. For commands and client-specific steps, [choose your AI app](connect/README.md).
To manage access after setup, see [Agents and connections](agents.md).

## Try your first task

Send your agent this prompt:

> Use Rukou to check which workspace you are connected to and whether browser sessions can
> start. Then start a session called first-visit, open https://example.com, and tell me the
> page title and one sentence about it. Stop the session when you finish.

The agent can use `rukou_status` to check its workspace and whether browsers can start. It
then starts a session, opens the page, reads it, and stops the browser. The
[MCP example](mcp-tools.md#complete-a-first-session) shows these calls explicitly.

## Watch and finish

Open **Sessions** in the same workspace and select `first-visit`. While it runs, you can watch
the browser. For a task that needs your input, use **Take control**, complete the step, and
select **Release** so the agent can continue.

When the task is finished, confirm the session has ended. You can also end a running session
through its menu with **Stop session**. Closing your AI app or the Rukou page does not stop
the session; it keeps running until it is stopped or reaches a time limit.

You are ready to give your agent a browsing task of your own. Add a [login](logins.md) when it needs
access to a website account, or use [troubleshooting](troubleshooting.md) if the first task fails.