# Your First Mission

You've got sandboxed.sh running. Now let's use it.

## Start a Mission

1. Open your dashboard at `https://agent.yourdomain.com`
2. Log in with your password
3. Click **New Mission** (or press `N`)
4. Type a simple task:

```
List the files in /root and tell me what you see.
```

5. Press Enter and watch

## What You'll See

The mission view shows:

- **Message stream**: Your messages and the agent's responses
- **Tool calls**: Every command the agent runs (with expandable output)
- **Thinking**: The agent's reasoning (if the model supports it)
- **Cost**: Running total of API usage

The agent will execute `ls -la /root`, read the output, and explain what it found.

## Try Something More Interesting

Here are good first tasks to understand capabilities:

**Code exploration:**
```
Clone https://github.com/Th0rgal/sandboxed.sh and summarize
what the main components do. Focus on src/api/.
```

**System administration:**
```
Check disk usage, memory, and running services on this server.
Highlight anything that looks unusual.
```

**File operations:**
```
Create a Python script in /tmp that fetches weather for Paris
and prints it. Use the wttr.in API. Then run it.
```

## Understanding the Output

### Tool Calls

Every action the agent takes appears as a tool call:

```
[Tool: bash]
Input: ls -la /root
Output: total 48
drwx------ 8 root root 4096 Jan 15 10:30 .
...
```

Click to expand and see full input/output.

### Mission Status

Missions have these states:

| Status | Meaning |
|--------|---------|
| `active` | Agent is working |
| `completed` | Agent finished successfully |
| `failed` | Something went wrong |
| `interrupted` | You cancelled it |

## Canceling a Mission

If the agent goes off track or you want to stop it:

1. Click **Cancel** in the top right
2. Or press `Esc`

The agent will stop after its current tool call completes.

## Workspaces

By default, missions run in the **host workspace**, directly on your server with full access.

For isolation, create a **container workspace**:

1. Go to **Workspaces** in the sidebar
2. Click **New Workspace**
3. Choose a template or configure manually
4. Select the workspace when starting a new mission

Container workspaces run in isolated Linux environments (systemd-nspawn). The agent can't affect your host system.

→ Learn more in [Workspaces](/workspaces)

## Next Steps

- **[Library](/library)**: Customize agent behavior with skills and rules
- **[Workspaces](/workspaces)**: Create isolated environments for different projects
- **[API Reference](/api)**: Integrate sandboxed.sh into your own tools