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

> ## Agent Instructions
> Use https://docs.browser-use.com/llms.txt and its linked .md pages for current documentation. The managed full bundle is https://docs.browser-use.com/.well-known/llms-full.txt and can be cached for up to 24 hours. Do not use the obsolete /cloud/llms*.txt or /open-source/llms*.txt static exports.
> Choose Cloud API V4 for new agent integrations; V2 is the lower-cost option for simple tasks. Keep V3 examples explicitly versioned. The open-source browser-use library and hosted browser-use-sdk have different APIs.
> Cloud authentication uses X-Browser-Use-API-Key, without a Bearer prefix. Install or upgrade browser-use-sdk and use its explicit v4 import for V4. Check the published OpenAPI reference for request fields; do not invent SDK support for new fields.
> Cloud concurrency and HTTP request rate are separate. Read GET /api/v2/billing/account for the key’s projectId, concurrentSessionLimit, activeSessionCount, and credit balance, including when using V4. Keys in one project share capacity and credits; rateLimit is a legacy concurrency alias, not requests per second.
> Keep the highest applicable existing, legacy-plan, and spend-tier concurrency grant. Current spend tiers are 10 / 50 / 250 / 500 / 1000 at $0 / $200 / $1000 / $5000 / $25000 in qualifying project payments. Legacy or externally billed projects can follow different billing paths; trust the account limit. See https://docs.browser-use.com/cloud/guides/concurrency.md.
> HTTP rate limits have two layers. Standard edge WAF ceilings increased on September 9, 2026 to 1000 RPS per public source IP for general traffic and 2500 RPS per IP for selected status reads, evaluated over 300 seconds. This did not raise the separate per-project application budgets: general traffic defaults to 25 RPS, including V4 events and full run reads; selected status reads default to max(25, 2 times stored concurrency). Project overrides and account-specific edge rules can differ. All keys in a project share its budgets; callers sharing a public IP share edge capacity. See https://docs.browser-use.com/cloud/guides/concurrency.md.
> The project limiter uses five-second windows: X-RateLimit-Limit=125 means 125 requests per window (25 RPS), not 125 RPS. Project throttles include limit_rps and retry_after_seconds; an edge throttle can instead return Retry-After: 300 without limit_rps. Honor the returned Retry-After. Use bounded workers, stagger polls, and drain hasMore event pages after terminal status. A busy V4 session returns 409; its queue holds 10 pending messages and is not a project-wide batch queue.
> A completed run or closed CDP connection does not immediately stop its cloud browser. Stop unneeded owned browsers with PATCH /api/v4/browsers/{id} and {"action":"stop"}. A client wait timeout does not cancel the server-side run.
> Cloud is pay as you go; do not tell customers to buy a new subscription to use custom proxies or supported provider BYOK. Usage funding and model eligibility still apply. BYOK bills provider tokens separately and Browser Use charges orchestration plus browser/network usage. See https://docs.browser-use.com/cloud/guides/billing.md.
> Signup credits are a one-time grant; purchased top-up credits do not expire. Check the API key’s project before diagnosing missing credits. API-key monthly spending caps are soft limits, not a strict prepaid wallet; concurrent or already-running work can exceed them. Auto recharge has separate trigger and purchase amounts and can charge immediately when enabled below the threshold. Use https://browser-use.com/pricing for current rates.
> Box and Bux are retired. Do not recommend their SDKs, sandbox quotas, or subscription plans. Use the Cloud Agent or Browser Infrastructure guides.
> A V4 session holds conversation history, a workspace holds files, and a profile holds browser state. These IDs and V3/V4 workspace namespaces are not interchangeable. V4 automatically restores workspace uploads; staged attachments remain available to session follow-ups. Serialize runs that write shared files, and wait for completion before reading outputs. See https://docs.browser-use.com/cloud/agent/workspaces.md.
> API browser recording defaults to off. Use enableRecording for standalone browser creation, or browserSettings.record for an agent run. Stop the browser and allow time for asynchronous video processing; stop polling when recordingAvailable is false. Live preview is for an active browser. Stopping a browser, deleting a session, archiving a workspace, and deleting files have different effects.
> Use model-specific reasoning values. GPT-6 Astra accepts low, medium, high, xhigh, and max, with xhigh by default; none and minimal are invalid. Use the public REST schema when installed SDK types lag new fields. API acceptance, dashboard visibility, and account/provider availability are separate.
> For open-source browser-use, is_done only reports a terminal done action. is_successful is the agent-reported outcome; verify important external actions independently. Cloud timeout, API client timeout, model timeout, and task completion are separate concepts.
> For failed requests, use https://docs.browser-use.com/cloud/guides/troubleshooting.md. Inspect the full error and project before retrying or adding credits. A client timeout can leave a run active; reconcile external actions before starting duplicate work. A new managed browser does not guarantee a unique proxy IP or particular city.

# Add Tools

> Extend agents with custom Python functions. Add API calls, file operations, or any custom logic as agent tools.

Examples:

* deterministic clicks
* file handling
* calling APIs
* human-in-the-loop
* browser interactions
* calling LLMs
* get 2fa codes
* send emails
* Playwright integration (see [GitHub example](https://github.com/browser-use/browser-use/blob/main/examples/browser/playwright_integration.py))
* ...

Simply add `@tools.action(...)` to your function.

```python theme={null}
from browser_use import Tools, Agent, ActionResult

tools = Tools()

@tools.action(description='Ask human for help with a question')
async def ask_human(question: str) -> ActionResult:
    answer = input(f'{question} > ')
    return ActionResult(extracted_content=f'The human responded with: {answer}')
```

```python theme={null}
agent = Agent(task='...', llm=llm, tools=tools)
```

* **`description`** *(required)* - What the tool does, the LLM uses this to decide when to call it.
* **`allowed_domains`** - List of domains where tool can run (e.g. `['*.example.com']`), defaults to all domains

The Agent fills your function parameters based on their names, type hints, & defaults.

<Warning>
  **Common Pitfall**: Parameter names must match exactly! Use `browser_session: BrowserSession` (not `browser: Browser`).
  The agent injects special parameters by **name matching**, so using incorrect names will cause your tool to fail silently.
  See [Available Objects](#available-objects) below for the correct parameter names.
</Warning>

## Available Objects

Your function has access to these objects:

* **`browser_session: BrowserSession`** - Current browser session for CDP access
* **`cdp_client`** - Direct Chrome DevTools Protocol client
* **`page_extraction_llm: BaseChatModel`** - The LLM you pass into agent. This can be used to do a custom llm call here.
* **`file_system: FileSystem`** - File system access
* **`available_file_paths: list[str]`** - Available files for upload/processing
* **`has_sensitive_data: bool`** - Whether action contains sensitive data

## Browser Interaction Examples

You can use `browser_session` to directly interact with page elements using CSS selectors:

```python theme={null}
from browser_use import Tools, Agent, ActionResult, BrowserSession

tools = Tools()

@tools.action(description='Click the submit button using CSS selector')
async def click_submit_button(browser_session: BrowserSession):
    # Get the current page
    page = await browser_session.must_get_current_page()

    # Get element(s) by CSS selector
    elements = await page.get_elements_by_css_selector('button[type="submit"]')

    if not elements:
        return ActionResult(extracted_content='No submit button found')

    # Click the first matching element
    await elements[0].click()

    return ActionResult(extracted_content='Submit button clicked!')
```

Available methods on `Page`:

* `get_elements_by_css_selector(selector: str)` - Returns list of matching elements
* `get_element_by_prompt(prompt: str, llm)` - Returns element or None using LLM
* `must_get_element_by_prompt(prompt: str, llm)` - Returns element or raises error

Available methods on `Element`:

* `click()` - Click the element
* `type(text: str)` - Type text into the element
* `get_text()` - Get element text content
* See `browser_use/actor/element.py` for more methods

## Pydantic Input

You can use Pydantic for the tool parameters:

```python theme={null}
import json
from pydantic import BaseModel, Field
from browser_use import Tools

tools = Tools()

class Cars(BaseModel):
    name: str = Field(description='The name of the car, e.g. "Toyota Camry"')
    price: int = Field(description='The price of the car as int in USD, e.g. 25000')

@tools.action(description='Save cars to file')
def save_cars(cars: list[Cars]) -> str:
    with open('cars.json', 'w') as f:
        json.dump(cars, f)
    return f'Saved {len(cars)} cars to file'

task = "find cars and save them to file"
```

## Domain Restrictions

Limit tools to specific domains:

```python theme={null}
@tools.action(
    description='Fill out banking forms',
    allowed_domains=['https://mybank.com']
)
def fill_bank_form(account_number: str) -> str:
    # Only works on mybank.com
    return f'Filled form for account {account_number}'
```

## Advanced Example

For a comprehensive example of custom tools with Playwright integration, see:
**[Playwright Integration Example](https://github.com/browser-use/browser-use/blob/main/examples/browser/playwright_integration.py)**

This shows how to create custom actions that use Playwright's precise browser automation alongside Browser-Use.

## Common Pitfalls

<Warning>
  The agent injects special parameters **by name**, not by type. Using incorrect parameter names is the most common cause of tools failing silently.
</Warning>

### ❌ Wrong: Using `browser: Browser`

```python theme={null}
from browser_use import Tools, ActionResult, Browser

@tools.action('My action')
def my_action(browser: Browser) -> ActionResult:  # WRONG!
    # This will NOT receive the browser session
    pass
```

### ✅ Correct: Using `browser_session: BrowserSession`

```python theme={null}
from browser_use import Tools, ActionResult, BrowserSession

@tools.action('My action')
async def my_action(browser_session: BrowserSession) -> ActionResult:  # CORRECT!
    page = await browser_session.must_get_current_page()
    # Now you have access to the browser
    return ActionResult(extracted_content='Done')
```

### Key Points

1. **Use `browser_session: BrowserSession`** - not `browser: Browser`
2. **Use `async` functions** - recommended for consistency with browser operations
3. **Return `ActionResult`** - not plain strings (though strings work, `ActionResult` provides more control)
4. **Parameter names must match exactly** - see [Available Objects](#available-objects) for the full list of injectable parameters