> ## Documentation Index
> Fetch the complete documentation index at: https://docs.browseros.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Your first run

> A guided walkthrough from a fresh BrowserClaw install to your first cockpit activity row.

Fresh install done. Open a new tab. This is the walkthrough for what you see next.

## What you will see

The cockpit greets you with a short motion demo, a single primary CTA to set up the MCP endpoint, a copyable starter prompt, and a three-step reminder strip.

Two visual states, depending on where you are:

* **First run**: no MCP connections, no activity. The MCP CTA pulses. Step 01 highlights.
* **Waiting**: MCP is connected, no activity yet. The CTA relabels to **View MCP endpoint**. Step 01 marks done. Step 02 highlights. A live banner tells you the cockpit is listening.

## Step 1: install the MCP into your agent

Click **Set up MCP endpoint**. You will land on the MCP install board.

Find your agent's row (Claude Code, Cursor, Codex, VS Code, Zed, OpenCode, Antigravity) and click **Install**. BrowserClaw writes the server URL into your agent's config file. Restart the agent.

<Card title="Full setup per harness" icon="plug" href="/browserclaw/mcp/index">
  Step-by-step for every supported client, plus the manual URL install for anything else.
</Card>

## Step 2: paste the starter prompt

Copy this into your agent and press enter:

```text theme={null}
Use BrowserClaw. Book me the cheapest morning flight from SFO to NYC next Friday.
```

Why this prompt: it exercises multi-tab navigation, form input, and result comparison. Three of BrowserClaw's most-used tool paths in one task. You will see all of them light up in the cockpit.

<Tip>
  The cockpit page has a Copy button next to this exact prompt. You do not need to type it out.
</Tip>

## Step 3: watch it land

Come back to the BrowserClaw new tab. Within seconds of your agent starting the task, a row appears under **Recent activity**:

* Left: the agent's name (`claude-code`, `cursor`, and so on).
* Middle: what it is doing (`browsing sfo to nyc, morning`).
* Right: a live status dot and a `just now` timestamp.

Click the row to open the task detail. You can watch the agent's timeline while it works, or replay any past step frame by frame.

## What happens after the first run

Once the cockpit has one activity row, the onboarding block disappears. From then on, every new tab lands on the populated cockpit: **Running now** at the top for live agents, **Recent activity** below for finished sessions.

## Next

<CardGroup cols={2}>
  <Card title="Sample prompts" icon="sparkles" href="/browserclaw/sample-prompts">
    A library of prompts across travel, research, workflow, and dev tasks.
  </Card>

  <Card title="The cockpit" icon="binoculars" href="/browserclaw/cockpit">
    Every region on the dashboard and what it tells you.
  </Card>
</CardGroup>