What it does
Launches a headless Chromium browser that the assistant can control — navigate to URLs, click elements, fill forms, read page content, and wait for changes. Uses accessibility snapshots to give the LLM numbered element references instead of fragile CSS selectors.
Requirements
Install Playwright and Chromium (one-time):
bunx playwright install chromium
How the assistant uses it
The browser tool exposes these actions:
| Action | What it does |
|---|
navigate | Go to a URL |
click | Click an element by its ref number |
type | Type text into an input field |
snapshot | Get the current page structure as numbered elements |
wait | Wait for a selector, URL pattern, page state, or fixed delay |
- Assistant calls
navigate to open a page
- Assistant calls
snapshot to see the page structure
- Assistant identifies the right element by its ref number
- Assistant calls
click or type to interact
Configuration
| Setting | Default | Description |
|---|
headless | true | Run without a visible browser window |
maxResultChars | 50,000 | Max characters returned from snapshots |
defaultTimeout | 30,000 ms | Timeout for page actions |
The browser tool is heavier than web fetch. The assistant prefers web_fetch for simple page reads and only launches the browser when it needs to interact with a page (click, fill forms, handle JavaScript-rendered content).
When to use browser vs web fetch
| Use browser when | Use web fetch when |
|---|
| Page requires JavaScript to render | Page is server-rendered HTML |
| You need to click or fill forms | You just need the text content |
| Content is behind interactions | Content is at a direct URL |
| You need to navigate through pages | Single page read is enough |
