> ## 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
> Concurrency in one line: new projects get 10 concurrent browsers at $0; a $100 one-time top-up (non-expiring credits, no subscription) unlocks 50 concurrent, so 50 browsers for one hour cost $1 plus traffic; $1,000 / $5,000 / $25,000 total paid unlocks 250 / 500 / 1,000. Details below and at https://docs.browser-use.com/cloud/guides/concurrency.md.
> 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; V3 trades accuracy for speed and lower cost than V4; V2 is legacy and only for existing integrations. Keep V3 examples explicitly versioned. The open-source browser-use library and hosted browser-use-sdk have different APIs.
> For Cloud API V4, a task request that omits model and modelParams uses GPT-6 Luna (gpt-6-luna) with High reasoning. Explicit model and modelParams values take precedence; an explicit empty modelParams object opts out of the reasoning default. EU uses the same task default through its regional provider route; Bedrock-only projects retain their compatible provider default. For latency-sensitive tasks set model to bu-ultrafast (bu-fast for more reasoning per step); these presets reject modelParams. See https://docs.browser-use.com/cloud/agent/ultrafast.md.
> 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 / $100 / $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. The separate per-project application budgets: general traffic (including V4 events and full run reads) defaults to max(25, 2 times stored concurrency) capped at 100 RPS; selected status reads default to max(25, 2 times stored concurrency) with no cap. The two are counted independently. 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 20 pending messages and is not a project-wide batch queue.
> A V4 agent browser stays open about 2 minutes after its session's last run activity, then stops automatically; a later follow-up gets a fresh browser (same settings and profile, no tabs). Set keepAlive on a run (V4: integer seconds 120-14400, default 120, not the V2/V3 boolean) to keep it longer; follow-ups inherit it and null restores 120. Free-tier and ZDR runs always use 120. Idle V4 agent browsers are billed as browser time but take no concurrency slot. A follow-up that reuses the open browser emits browser.reattached instead of browser.ready. Closing CDP does not stop a 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 1.0, result.is_done() means the run completed (status completed); result.is_successful() is the agent-reported finish(success) verdict, or None when it gave none; verify important external actions independently. Cloud timeout, API client timeout, model timeout, and task completion are separate concepts.
> Pages under /open-source/legacy/0.x/ document browser-use 0.x (pip install 'browser-use<1'). For new open-source code, use the 1.0 pages under /open-source/ outside legacy/.
> 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.
> V4 secretBindings.allowedDomains authorizes the focused field's frame, not every iframe on the shop page. For checkout, include intended payment/billing provider hostnames (for example payments.bigcommerce.com). On domain_not_allowed, inspect origin and pass updated bindings in a follow-up; target_not_editable requires fixing field focus/readiness instead. See https://docs.browser-use.com/cloud/guides/secrets.md.
> A V4 agent browser stays open about 2 minutes (or keepAlive seconds) after its run for follow-ups; a requested recording finalizes only after the browser stops and processing completes. Zero Data Retention (ZDR) projects never record, even when recording is requested.
> If you hit an issue you cannot explain, email contact@browser-use.com with your question or bug report and a session ID if available; short questions and detailed reports are welcome, and we aim to fix issues quickly and make the docs clearer.

# Migrating from 0.x

> What changed in browser-use 1.0 and why, before and after code for the ten most common 0.x patterns, every removed option with its replacement, and how to stay on 0.x.

1.0 replaces the engine. The agent no longer picks one action per step from a DOM dump: it writes Python in a persistent REPL with [browser-harness](https://github.com/browser-use/browser-harness) helpers and raw CDP, on [Pydantic AI](https://ai.pydantic.dev). The common surface (`Agent`, `Browser`, `Tools`, `run()`, `final_result()`) still works.

Anything 1.0 can't honor raises a clear error before a browser starts or anything is billed, never a silent no-op. A removed feature raises `<feature> was removed in browser-use 1.0`, plus the 1.0 alternative when there is one and a link to the migration guide. The exception: a few tuning knobs that can't change the result (`use_vision=True`, `flash_mode`, `use_thinking`, ...) log one warning and are ignored.

## Stay on 0.x

Pin it:

```bash theme={"theme":"css-variables"}
pip install 'browser-use<1'
```

With uv: `uv add 'browser-use<1'`. In `requirements.txt` or `pyproject.toml`: `browser-use<1`. The 0.x docs are under [Legacy (0.x)](https://docs.browser-use.com/open-source/legacy/0.x/introduction.md).

## What changed and why

<picture>
  <source media="(max-width: 600px)" />

  ![In 0.x the model read an indexed DOM dump and picked one action, like click(index=12), for a local Chromium. In 1.0 the model reads the output of its last cell and writes the next Python cell, which the REPL runs with browser helpers and raw CDP against a cloud browser by default. Agent, Browser, Tools, run() and final\_result() keep working.](https://mintcdn.com/browseruse-0aece648/vou-GNVvaEF9jnyJ/open-source/images/oss-migrating-light.svg?fit=max\&auto=format\&n=vou-GNVvaEF9jnyJ\&q=85\&s=c6424215abf7e48582cfb35acb296b23)
</picture>

| | 0.x | 1.0 | Why |
| - | - | - | - |
| Each step | The model reads an indexed DOM dump and picks actions like `click(index=12)`. | The model writes one Python cell; a persistent REPL runs it with browser helpers and raw CDP. | One cell can click, type, read and loop. Anything Chrome DevTools can do, the agent can do, without a built-in action for it. |
| Agent loop | browser-use's own | [Pydantic AI](https://ai.pydantic.dev) | Any Pydantic AI model works, and you get its OpenTelemetry traces. |
| Model (`llm=None`) | required | `bu-fast` on the Browser Use gateway | The configuration we evaluate. No provider account needed. |
| Browser (`Browser()`) | local Chromium | a Browser Use Cloud browser | Stealth, proxies and CAPTCHA solving, nothing to install. A local browser is `Browser('chromium')`. |
| Key | one per provider | one Browser Use key: `BROWSER_USE_API_KEY` (or `browser-use login`) | One key runs the model, cloud browsers, web search, the CLI and the MCP server. |
| `run(max_steps=)` | 500 | 100 | One step is now one model request, and a cell does more than one 0.x action. |
| A removed option | often ignored | raises before a browser starts | You find out before anything is billed. |
| Tasks with localhost / private IPs | worked | a cloud browser can't reach them: one warning; use `Browser('chromium')` | |
| Cloud proxy | the API's default, `us` | `us`; `cloud_proxy_country_code=None` turns it off | |

## Common errors

* **`MissingAPIKeyError: The default model (bu-fast) needs a Browser Use API key`** (or `Browser() (a Browser Use Cloud browser) needs a Browser Use API key`): the default model and `Browser()` are Browser Use Cloud now. Set `BROWSER_USE_API_KEY` (create a key at [cloud.browser-use.com/new-api-key](https://cloud.browser-use.com/new-api-key)) or run `browser-use login`, or stay local: `Agent(task, llm=ChatOpenAI(...), browser=Browser('chromium'))`.
* **`ValueError: Browser() (a cloud browser) does not take headless=`**: `headless`, `executable_path` and `user_data_dir` are for local browsers. Use `Browser('chromium', headless=True)`.
* **`TypeError: Agent(<option>=...) was removed in browser-use 1.0`**: an option 1.0 dropped. The message names the replacement when there is one; the tables below list them all.
* **`TypeError: Agent(max_steps=...) is not an option`**: pass it to `agent.run(max_steps=...)`.
* **`ValueError: bu-2-0 needs browser-use<1.0`**: the 0.x gateway models are gone. Use `bu-fast`, or stay on 0.x.
* **`BrowserConfigError: No Chrome profiles match 'Gaming'. Your profiles: ...`**: `profile=` (or `cloud_profile_id=`) names no profile. It is a `ValueError`, raised from `run()` before a browser starts.
* **`ModuleNotFoundError` / `ImportError` on a 0.x import**: the message names the 1.0 replacement. Import from the top level: `from browser_use import Agent, Browser, Tools, ActionResult, ChatOpenAI, ...`.

## The ten most common 0.x patterns

The 0.x snippets are written for browser-use 0.13.14. Run as is on 1.0, each one either works or raises before a browser starts, naming the fix.

### 1. Your own model

0.x launched a local browser by default. 1.0 starts a cloud browser, so add `Browser('chromium')` to stay local. `max_steps` still goes to `run()`.

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, ChatOpenAI


async def main():
    agent = Agent(task='Find the top story on Hacker News', llm=ChatOpenAI(model='gpt-4.1-mini'))
    history = await agent.run(max_steps=20)
    print(history.final_result())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from browser_use import Agent, Browser, ChatOpenAI

agent = Agent(
    'Find the top story on Hacker News',
    llm=ChatOpenAI('gpt-6-luna'),
    browser=Browser('chromium'),  # 0.x launched a local browser by default; 1.0's default is a cloud browser
)
result = agent.run_sync(max_steps=20)
print(result.final_result())
```

### 2. The Browser Use model

`ChatBrowserUse()` meant `bu-2-0` on the 0.x gateway. On 1.0 it is `bu-fast` on the new gateway, and it is the default, so you can drop `llm=`. `bu-2-0`, `bu-1-0`, `bu-2-0-mini-preview` and `bu-latest` raise: they need 0.x.

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, ChatBrowserUse


async def main():
    agent = Agent(task='Find the top story on Hacker News', llm=ChatBrowserUse())  # bu-2-0
    history = await agent.run()
    print(history.final_result())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from browser_use import Agent

result = Agent('Find the top story on Hacker News').run_sync()  # bu-fast on a cloud browser, one Browser Use key
print(result.final_result())
```

### 3. BrowserProfile and BrowserSession

Pass the options to `Browser(...)`. `headless`, `user_data_dir`, `executable_path` and the domain lists are kept. Window size, viewport and user agent are said in the task.

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, BrowserProfile, BrowserSession, ChatOpenAI


async def main():
    profile = BrowserProfile(
        headless=True,
        user_data_dir='~/.config/browseruse/profiles/work',
        window_size={'width': 1280, 'height': 800},
        allowed_domains=['example.com'],
    )
    agent = Agent(
        task='What is the title of https://example.com?',
        llm=ChatOpenAI(model='gpt-4.1-mini'),
        browser_session=BrowserSession(browser_profile=profile),
    )
    history = await agent.run()
    print(history.final_result())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from browser_use import Agent, Browser, ChatOpenAI

browser = Browser(
    'chromium',
    headless=True,
    user_data_dir='~/.config/browseruse/profiles/work',
    allowed_domains=['example.com'],
)
agent = Agent(
    'What is the title of https://example.com?',  # window_size: say it in the task if the page needs it
    llm=ChatOpenAI('gpt-6-luna'),
    browser=browser,
)
print(agent.run_sync().final_result())
```

### 4. Your Chrome profile

`Browser('chrome', profile='Default')` launches a new Chrome on a temporary copy of that profile (its display name or directory): your logins, sync off, nothing written back. `Browser('chrome')` attaches to the Chrome you have open instead (turn on `chrome://inspect/#remote-debugging` once and click **Allow**). See [Browsers](https://docs.browser-use.com/open-source/browsers.md#your-chrome).

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, Browser, ChatOpenAI


async def main():
    browser = Browser(
        executable_path='/Applications/Google Chrome.app/Contents/MacOS/Google Chrome',
        user_data_dir='~/Library/Application Support/Google/Chrome',
        profile_directory='Default',
    )
    agent = Agent(
        task='What is the title of https://example.com?', llm=ChatOpenAI(model='gpt-4.1-mini'), browser=browser
    )
    history = await agent.run()
    print(history.final_result())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from browser_use import Agent, Browser, ChatOpenAI

browser = Browser('chrome', profile='Default')  # a new Chrome on a temporary copy of that profile
agent = Agent('What is the title of https://example.com?', llm=ChatOpenAI('gpt-6-luna'), browser=browser)
print(agent.run_sync().final_result())
```

### 5. Cloud browsers

`Browser()` is a cloud browser now; `use_cloud=True` still works. `cloud_profile_id`, `cloud_proxy_country_code` and `cloud_timeout` are kept.

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, Browser, ChatOpenAI


async def main():
    browser = Browser(use_cloud=True, cloud_proxy_country_code='de', cloud_timeout=15)
    agent = Agent(
        task='Open https://ipinfo.io/json in the browser and tell me which country it shows',
        llm=ChatOpenAI(model='gpt-4.1-mini'),
        browser=browser,
    )
    history = await agent.run()
    print(history.final_result())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from browser_use import Agent, Browser, ChatOpenAI

browser = Browser(cloud_proxy_country_code='de', cloud_timeout=15)  # use_cloud=True is the default now
agent = Agent(
    'Open https://ipinfo.io/json in the browser and tell me which country it shows',
    llm=ChatOpenAI('gpt-6-luna'),
    browser=browser,
)
print(agent.run_sync().final_result())
```

### 6. Custom tools

`@tools.action` is kept. Injected parameters (`browser_session`, `page_extraction_llm`, `file_system`, ...) are gone: a tool gets only what the model passes, and the model drives the browser itself. Return a string, any JSON-able value or `ActionResult(extracted_content=...)`.

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import ActionResult, Agent, BrowserSession, ChatOpenAI, Tools

tools = Tools()


@tools.action('Save a note about the current page')
async def save_note(note: str, browser_session: BrowserSession) -> ActionResult:
    url = await browser_session.get_current_page_url()
    with open('notes.txt', 'a') as f:
        f.write(f'{url}: {note}\n')
    return ActionResult(extracted_content='saved', include_in_memory=True)


async def main():
    agent = Agent(
        task='Open https://example.com and save a note about what the page says',
        llm=ChatOpenAI(model='gpt-4.1-mini'),
        tools=tools,
    )
    await agent.run()
    print(open('notes.txt').read())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from browser_use import Agent, ChatOpenAI, Tools

tools = Tools()


@tools.action('Save a note about a page. Pass the page URL and the note.')
def save_note(url: str, note: str) -> str:
    with open('notes.txt', 'a') as f:
        f.write(f'{url}: {note}\n')
    return 'saved'


agent = Agent(
    'Open https://example.com and save a note about what the page says',
    llm=ChatOpenAI('gpt-6-luna'),
    tools=tools,
)
agent.run_sync()
print(open('notes.txt').read())
```

### 7. Structured output

`output_model_schema` is kept. Read the typed instance from `result.structured_output` instead of parsing `final_result()`.

```python 0.x theme={"theme":"css-variables"}
import asyncio

from pydantic import BaseModel

from browser_use import Agent, ChatOpenAI


class Story(BaseModel):
    title: str
    points: int


class Stories(BaseModel):
    stories: list[Story]


async def main():
    agent = Agent(
        task='Top 3 stories on https://news.ycombinator.com',
        llm=ChatOpenAI(model='gpt-4.1-mini'),
        output_model_schema=Stories,
    )
    history = await agent.run()
    stories = Stories.model_validate_json(history.final_result())
    for s in stories.stories:
        print(s.points, s.title)


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from pydantic import BaseModel

from browser_use import Agent, ChatOpenAI


class Story(BaseModel):
    title: str
    points: int


class Stories(BaseModel):
    stories: list[Story]


agent = Agent(
    'Top 3 stories on https://news.ycombinator.com', llm=ChatOpenAI('gpt-6-luna'), output_model_schema=Stories
)
result = agent.run_sync()
stories = result.structured_output  # a Stories instance; None if the run did not complete
if stories:
    for s in stories.stories:
        print(s.points, s.title)
```

### 8. Logins with sensitive\_data

Unchanged. The model types a value with `type_secret('x_pass')` and never sees it. The 0.x `<secret>x_pass</secret>` placeholder is no longer replaced: code that sends it to a page raises, naming `type_secret`. See [Secrets](https://docs.browser-use.com/open-source/secrets.md).

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, ChatOpenAI


async def main():
    agent = Agent(
        task='Log in on https://the-internet.herokuapp.com/login with x_user and x_pass, then return the message the page shows',
        llm=ChatOpenAI(model='gpt-4.1-mini'),
        sensitive_data={'https://the-internet.herokuapp.com': {'x_user': 'tomsmith', 'x_pass': 'SuperSecretPassword!'}},
        use_vision=False,
    )
    history = await agent.run()
    print(history.final_result())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
from browser_use import Agent, ChatOpenAI

agent = Agent(
    'Log in on https://the-internet.herokuapp.com/login with x_user and x_pass, then return the message the page shows',
    llm=ChatOpenAI('gpt-6-luna'),
    sensitive_data={'https://the-internet.herokuapp.com': {'x_user': 'tomsmith', 'x_pass': 'SuperSecretPassword!'}},
    use_vision=False,  # kept: no screenshot reaches the model
)
print(agent.run_sync().final_result())
```

### 9. Follow-up tasks on one browser

`keep_alive=True` and `add_new_task()` are kept. Stop the browser with `await browser.stop()`; `kill()` and `close()` do the same.

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, Browser, ChatOpenAI


async def main():
    browser = Browser(keep_alive=True)
    agent = Agent(
        task='Open https://news.ycombinator.com and return the top story title',
        llm=ChatOpenAI(model='gpt-4.1-mini'),
        browser=browser,
    )
    print((await agent.run()).final_result())
    agent.add_new_task('Now open its comments and return the number of comments')
    print((await agent.run()).final_result())
    await browser.kill()


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, Browser, ChatOpenAI


async def main():
    browser = Browser(keep_alive=True)
    agent = Agent(
        'Open https://news.ycombinator.com and return the top story title',
        llm=ChatOpenAI('gpt-6-luna'),
        browser=browser,
    )
    try:
        print((await agent.run()).final_result())
        agent.add_new_task('Now open its comments and return the number of comments')
        print((await agent.run()).final_result())
    finally:
        await browser.stop()  # browser.kill() and browser.close() also work


asyncio.run(main())
```

### 10. Step hooks and history

There are no per-step actions in 1.0, so `on_step_start` / `on_step_end` and `history.urls()`, `action_names()` and the like are gone. Each step logs one line to `logging.getLogger('browser_use')`, and `result.messages` holds the full transcript. See [Long tasks: Live progress](https://docs.browser-use.com/open-source/long-tasks.md#live-progress).

```python 0.x theme={"theme":"css-variables"}
import asyncio

from browser_use import Agent, ChatOpenAI


async def on_step_end(agent):
    print('step', agent.state.n_steps, await agent.browser_session.get_current_page_url())


async def main():
    agent = Agent(task='What is the title of https://example.com?', llm=ChatOpenAI(model='gpt-4.1-mini'))
    history = await agent.run(max_steps=10, on_step_end=on_step_end)
    print(history.urls())
    print(history.action_names())
    print(history.number_of_steps(), history.total_duration_seconds())


asyncio.run(main())
```

```python 1.0 theme={"theme":"css-variables"}
import logging

from browser_use import Agent, ChatOpenAI


class Progress(logging.Handler):
    def emit(self, record):
        print('progress:', record.getMessage())  # one line per step: step N: <first line of the cell>


logging.getLogger('browser_use').addHandler(Progress())

agent = Agent('What is the title of https://example.com?', llm=ChatOpenAI('gpt-6-luna'))
result = agent.run_sync(max_steps=10)
print(result.number_of_steps(), result.total_duration_seconds(), result.cost_usd)
print(len(result.messages), 'messages in the transcript')
```

## Agent

Every 0.x `Agent(...)` argument, method and attribute. Removed arguments raise `TypeError`, removed methods and attributes raise `AttributeError`, both naming the 1.0 replacement. "Dropped" means there is none.

| 0.x | 1.0 |
| - | - |
| `task`, `llm`, `browser`, `browser_session`, `tools`, `controller`, `output_model_schema`, `sensitive_data`, `extend_system_message` | kept |
| `use_vision=False` | kept: no screenshot reaches the model |
| `use_vision=True` / `'auto'`, `vision_detail_level`, `llm_screenshot_size`, `max_actions_per_step`, `flash_mode`, `use_thinking`, `use_judge`, `max_failures`, `final_response_after_failure`, `calculate_cost`, `highlight_elements` | ignored, one warning each. Instead of `use_thinking`: `reasoning='low' / 'medium' / 'high'`; of `flash_mode`: `mode='ultrafast'`; of `highlight_elements`: `highlight_actions` |
| `max_steps` on `Agent` | not an option: `agent.run(max_steps=...)` |
| `override_system_message` | `extend_system_message` |
| `llm_timeout`, `step_timeout` | `agent.run(timeout=seconds)` |
| `file_system_path`, `available_file_paths` | `workspace=`: put the files in it and name them in the task. See [Files](https://docs.browser-use.com/open-source/files.md) |
| `initial_actions`, `directly_open_url` | say it in the task |
| `browser_profile`, `browser_context`, `context`, `page` | `browser=Browser(...)` |
| `page_extraction_llm`, `extraction_schema` | dropped: the agent extracts with Python |
| `register_new_step_callback`, `register_done_callback`, `register_external_agent_status_raise_error_callback`, `register_should_stop_callback`, `run(on_step_start=...)`, `run(on_step_end=...)` | a `logging.Handler` on `browser_use`; see [Long tasks](https://docs.browser-use.com/open-source/long-tasks.md#live-progress) |
| `save_conversation_path`, `save_conversation_path_encoding` | `result.messages` |
| `generate_gif`, `demo_mode` | dropped. `live_view=True` shows a cloud browser live |
| `judge_llm`, `ground_truth` | `result.success`, the agent's own verdict |
| `fallback_llm`, `skill_ids`, `skills`, `skill_service`, `pricing_url` | dropped |
| `sample_images`, `enable_planning`, `planning_replan_on_stall`, `planning_exploration_limit`, `planner_llm`, `planner_interval`, `is_planner_reasoning`, `extend_planner_system_message`, `use_vision_for_planner` | dropped: each step is one Python cell; the agent takes screenshots when it needs them |
| `loop_detection_window`, `loop_detection_enabled`, `message_compaction`, `max_history_items`, `include_attributes`, `max_clickable_elements_length`, `include_recent_events`, `include_tool_call_examples`, `display_files_in_done_text`, `injected_agent_state`, `source`, `task_id`, `enable_signal_handler`, `_url_shortening_limit` | dropped: the model retries failed cells within `max_steps`, and history is compacted on its own (`max_context_chars`) |
| 0.1 to 0.8: `validate_output`, `system_prompt_class`, `message_context`, `enable_memory`, `memory_config`, `memory_interval`, `max_input_tokens`, `images_per_step`, `retry_delay`, `save_playwright_script_path`, `cloud_sync`, `preload`, `max_error_length`, `tool_calling_method`, `tool_call_in_content` | dropped (`result.success` for `validate_output`, `extend_system_message` for the prompt ones) |
| `pause()`, `resume()`, `stop()`, `step()`, `take_step()`, `multi_act()` | cancel the task awaiting `run()`; see [Agent](https://docs.browser-use.com/open-source/agent.md#cancel). For a partial result, bound the run with `max_steps` and `timeout` |
| `save_history()`, `load_and_rerun()`, `rerun_history()`, `detect_variables()` | dropped: `result.messages` holds the transcript; run the task again |
| `set_task()` | `add_new_task(task)` |
| `wait_until_resumed()`, `get_next_action()`, `get_model_output()`, `set_tool_calling_method()`, `create_history_gif()`, `get_trace_object()`, `authenticate_cloud_sync()`, `save_file_system_state()`, `log_completion()` | dropped: 0.x loop internals |
| `logger` | `logging.getLogger('browser_use')` |
| `current_llm_model`, `is_using_fallback_llm` | `agent.llm` |
| `state`, `history`, `message_manager`, `settings`, `file_system` | dropped: read the result `run()` returns, and `agent.workspace` |
| `browser_session`, `browser_profile`, `browser_context`, `controller` (attributes) | `agent.browser`, `agent.tools` |
| new | `reasoning`, `mode`, `workspace`, `web_search`, `max_context_chars`, `highlight_actions`. See [Agent](https://docs.browser-use.com/open-source/agent.md) |

## Result

`run()` still returns an `AgentHistoryList`.

| 0.x | 1.0 |
| - | - |
| `final_result()`, `is_done()`, `structured_output`, `number_of_steps()`, `total_duration_seconds()` | kept |
| `is_successful()` | the agent's verdict, `None` when it gave none (0.x defaulted to `True`). For the 0.x behavior: `r.is_done() and r.success is not False` |
| `errors()`, `has_errors()` | only the error that ended the run. Cell errors the agent recovered from are in `result.messages` |
| `history`, `urls()`, `screenshots()`, `screenshot_paths()`, `action_names()`, `action_history()`, `action_results()`, `model_thoughts()`, `model_outputs()`, `model_actions()`, `model_actions_filtered()`, `extracted_content()`, `agent_steps()`, `last_action()`, `add_item()` | dropped (`AttributeError`): there are no per-step actions. Use `result.messages`, `result.partial` and the files in `agent.workspace` |
| `save_to_file()`, `load_from_file()`, `load_from_dict()`, `model_dump()`, `model_dump_json()`, `save_as_playwright_script()` | dropped |
| `get_structured_output()` | `result.structured_output` |
| `judgement()`, `is_judged()`, `is_validated()` | `result.success` |
| `input_token_usage()`, `total_input_tokens`, `result.usage.total_cost` | `result.usage['input']`, `result.cost_usd` |
| new | `status`, `success`, `partial`, `usage`, `cost_usd`, `messages`, `downloads`. See [Output](https://docs.browser-use.com/open-source/output.md) |

## Browser

`Browser`, `BrowserSession` and `BrowserProfile` are one class now: `Browser`. Importing `BrowserSession`, `BrowserProfile`, `BrowserConfig`, `BrowserContext` or `BrowserContextConfig` raises.

| 0.x | 1.0 |
| - | - |
| `Browser()`, `BrowserSession()` | `Browser()` is a cloud browser |
| `Browser(use_cloud=True)` | `Browser()`; `use_cloud=True` still works. `cloud_browser=True` raises: use `Browser()` |
| `Browser(use_cloud=False)` | `Browser('chromium')` (or `local=True`): a fresh local Chrome, Chromium or Edge with a temp profile |
| `headless`, `executable_path` | kept for browsers 1.0 launches; on `Browser()` they raise `ValueError` |
| `user_data_dir` | kept with `Browser('chromium', user_data_dir=...)`. Chrome's own default directory raises: Chrome 136+ refuses remote debugging on it |
| `Browser.from_system_chrome()`, `profile_directory` | `Browser('chrome')` (your open Chrome) or `Browser('chrome', profile='Work')` |
| `keep_alive`, `cdp_url` | kept |
| `cloud_profile_id`, `cloud_proxy_country_code`, `cloud_timeout` | kept. The aliases `profile_id`, `proxy_country_code` and `timeout` raise, naming these |
| `cloud_browser_params` | `cloud_profile_id`, `cloud_proxy_country_code`, `cloud_timeout` |
| `await browser.close()`, `await browser.kill()` | kept: they stop the browser, as `await browser.stop()` does |
| `stop_session()` | `await browser.stop()` |
| `allowed_domains`, `prohibited_domains`, `block_ip_addresses` | kept, in every mode; see [Domains](https://docs.browser-use.com/open-source/domains.md) |
| `captcha_solver` | kept for cloud browsers; `False` raises on a local one |
| `headers` (with `cdp_url`) | raises: use a `cdp_url` that needs no headers |
| `window_size`, `window_width`, `window_height`, `window_position`, `viewport`, `no_viewport`, `screen`, `device_scale_factor`, `user_agent` | say it in the task: the agent can send `Emulation.setUserAgentOverride` and, on a local browser, `Emulation.setDeviceMetricsOverride`. A cloud browser keeps its own 1440x900 viewport; use `Browser('chromium')` for another. Headless launches use a 1440x900 window |
| `locale`, `timezone_id`, `color_scheme`, `contrast`, `forced_colors`, `reduced_motion`, `is_mobile`, `has_touch`, `extra_http_headers`, `offline`, `java_script_enabled`, `bypass_csp`, `ignore_https_errors`, `service_workers`, `base_url` (0.6 Playwright options) | say it in the task: the agent sends the CDP commands itself |
| `permissions` | say it in the task: the agent can send `Browser.grantPermissions` |
| `proxy` | cloud: `cloud_proxy_country_code`. Local: dropped for now |
| `storage_state`, `cookie_whitelist_domains`, `cookies_file` | cloud: `cloud_profile_id`. Local: `Browser('chrome', profile=...)` or `Browser('chromium', user_data_dir=...)` |
| `downloads_path`, `accept_downloads`, `auto_download_pdfs` | downloads land in `agent.workspace/downloads`; see [Files](https://docs.browser-use.com/open-source/files.md#downloads) |
| `highlight_elements`, `interaction_highlight_color`, `interaction_highlight_duration` | actions are highlighted by default; `Agent(highlight_actions=False)` turns it off |
| `channel` | `executable_path=` |
| `args`, `ignore_default_args`, `env`, `chromium_sandbox`, `devtools`, `enable_default_extensions`, `disable_security`, `is_local`, `id`, `config`, `browser_profile`, `context` | dropped (`--no-sandbox` is added on Linux as root or in a container) |
| `record_video_dir`, `record_video_size`, `record_video_framerate`, `record_har_path`, `record_har_content`, `record_har_mode`, `record_har_omit_content`, `record_har_url_filter`, `traces_dir` | dropped. `result.messages` is the transcript; `live_view=True` shows a cloud browser live |
| `dom_highlight_elements`, `filter_highlight_ids`, `paint_order_filtering`, `cross_origin_iframes`, `max_iframes`, `max_iframe_depth`, `minimum_wait_page_load_time`, `wait_for_network_idle_page_load_time`, `maximum_wait_page_load_time`, `wait_between_actions`, `deterministic_rendering`, `demo_mode`, `viewport_expansion`, `include_dynamic_attributes`, `cached_state`, `default_timeout`, `default_navigation_timeout`, `slow_mo`, `strict_selectors` | dropped: they tuned the 0.x DOM engine |
| `handle_sigint`, `handle_sigterm`, `handle_sighup` | dropped: a browser 1.0 started is stopped when the process exits |
| `stealth` | dropped: cloud browsers run with stealth settings |
| `get_current_page()`, `new_page()`, `navigate_to()`, `navigate()`, `get_tabs()`, `switch_tab()`, `close_tab()`, `take_screenshot()`, `get_browser_state_summary()`, `get_page_html()`, `get_cookies()`, `export_storage_state()`, `cdp_client`, `event_bus`, and the other 0.x `BrowserSession` / `BrowserContext` methods | dropped (`AttributeError`): the agent drives pages in Python with browser helpers and raw CDP |
| `Browser.list_chrome_profiles()` | `Browser('chrome', profile='Work')` takes a display or directory name; one that matches none raises with your profiles listed |

See [Browsers](https://docs.browser-use.com/open-source/browsers.md).

## Models

| 0.x | 1.0 |
| - | - |
| `ChatBrowserUse()` | kept: `bu-fast` (default), `bu-ultrafast`, `bu-smart` or an OpenAI model the gateway lists. `'openai/gpt-5.5'` is sent as `gpt-5.5` |
| `ChatBrowserUse('bu-2-0')`, `'bu-1-0'`, `'bu-2-0-mini-preview'`, `'bu-latest'` | raise `ValueError`: they need `browser-use<1`. Use `bu-fast` |
| `ChatBrowserUse('anthropic/...')`, `ChatBrowserUse('google/...')` | the 0.x gateway served other vendors; the 1.0 error names `ChatAnthropic(...)` / `ChatGoogle(...)` with your own key |
| `ChatOpenAI(model, api_key, base_url, temperature, reasoning_effort)` | kept. api.openai.com uses the Responses API; any other `base_url` uses chat completions |
| `ChatAnthropic(model, api_key)`, `ChatGoogle(model, api_key)` | kept, built in. `ChatAnthropic(temperature=...)` raises: delete it (Anthropic rejects it on Claude 4.x) |
| other `ChatOpenAI` / `ChatAnthropic` / `ChatGoogle` arguments | raise `TypeError`: pass a Pydantic AI model built with `settings=...`, e.g. `GoogleModel('gemini-flash-latest', settings={'temperature': 0.5})` |
| `ChatAzureOpenAI` | `llm='azure:<deployment>'` with `AZURE_OPENAI_ENDPOINT`, `AZURE_OPENAI_API_KEY` (0.x read `AZURE_OPENAI_KEY`) and `OPENAI_API_VERSION` (now required) |
| `ChatAWSBedrock`, `ChatAnthropicBedrock` | `llm='bedrock:<model id>'` after `pip install 'pydantic-ai-slim[bedrock]'` |
| `ChatGroq`, `ChatMistral` | `llm='groq:<model>'` / `'mistral:<model>'` after `pip install 'pydantic-ai-slim[groq]'` (or `[mistral]`) |
| `ChatDeepSeek`, `ChatOpenRouter`, `ChatCerebras`, `ChatVercel` | `'deepseek:deepseek-chat'`, `'openrouter:<vendor>/<model>'`, `'cerebras:<model>'`, `'vercel:<vendor>/<model>'` |
| `ChatOllama` | `ChatOpenAI('<model>', base_url='http://localhost:11434/v1', api_key='ollama')` |
| `ChatLiteLLM`, `ChatOCIRaw`, `ChatOrcaRouter` | `ChatOpenAI(model, base_url=...)` for its OpenAI-compatible endpoint |
| `browser_use.llm.models`, `openai_gpt_4o` and the other 0.x model instances, `get_llm_by_name()` | a model string or Chat class; each error names it (`ChatOpenAI('gpt-4o')` for `openai_gpt_4o`) |
| `llm.ainvoke(...)`, `UserMessage`, `SystemMessage`, `BaseChatModel` | dropped: the agent calls the model. To call one yourself: `await pydantic_ai.Agent(llm.pydantic_model()).run(prompt)` |
| LangChain chat models | not supported: `Agent(llm=...)` raises `TypeError`. Use `from browser_use import ChatOpenAI` |
| a bare model name such as `llm='gpt-4o'` | raises naming its 1.0 form: `ChatOpenAI('gpt-4o')` or `'openai:gpt-4o'` |

Importing a removed class raises an `ImportError` that names its replacement. See [Models](https://docs.browser-use.com/open-source/models.md).

## Tools

| 0.x | 1.0 |
| - | - |
| `Tools()`, `Controller()`, `@tools.action('...')`, `param_model` | kept. `Controller` is `Tools` |
| returning `ActionResult(extracted_content=..., error=...)` or any value | kept |
| an action that raises | kept: the model gets the error and can recover |
| injected parameters (`browser_session`, `page_extraction_llm`, `file_system`, `available_file_paths`, `has_sensitive_data`, `cdp_client`, `page_url`, ...) | dropped: actions get only what the model passes |
| `@tools.action(domains=...)`, `allowed_domains=`, `page_filter=` | `Browser(allowed_domains=...)` |
| `@tools.action(requires_browser=...)`, `terminates_sequence=` | dropped: delete them |
| `Tools(exclude_actions=...)` | dropped: there are no built-in actions; the agent writes Python |
| `Tools(output_model=...)` | `Agent(output_model_schema=...)` |
| `Tools(display_files_in_done_text=...)`; 0.1 `Controller(headless=, keep_open=, disable_security=, cookies_path=, browser_config=)` | dropped; browser options go to `Browser(...)` |
| `act()`, `exclude_action()`, `get_output_model()`, `use_structured_output_action()`, `set_coordinate_clicking()` | dropped (`AttributeError` naming the answer) |
| `ActionResult(is_done=True)`, `success=`, `attachments=[...]`, `images=[...]` | dropped: the model ends the run with `finish`; save files in the workspace. An action that returns one raises `TypeError` when `@tools.action` registers it |
| `ActionResult(include_in_memory=...)`, `include_extracted_content_only_once=` | ignored, one warning: every result stays in the conversation |
| `<secret>name</secret>` in an action's arguments | dropped: the action is not run and the model gets an error. Secrets are typed in the browser with `type_secret(name)` |

See [Tools](https://docs.browser-use.com/open-source/tools.md).

## Removed modules

Every 0.x module 1.0 does not have, and every 0.x name a kept module lost, raises an error naming the replacement: `ModuleNotFoundError` for a module, `ImportError` for a name. Both are `ImportError`, so `except ImportError` fallbacks still work. Import from the top level: `from browser_use import Agent, Browser, Tools, ActionResult, ChatBrowserUse, ChatOpenAI, ChatAnthropic, ChatGoogle`.

| 0.x | 1.0 |
| - | - |
| `browser_use.llm.openai`, `.anthropic`, `.google`, `.browser_use` | the top-level `Chat*` classes |
| `browser_use.llm.aws`, `.azure`, `.deepseek`, `.groq`, `.mistral`, `.openrouter`, `.cerebras`, `.vercel`, `.ollama`, `.litellm`, `.oci_raw`, `.orcarouter` | the model strings in [Models](#models) |
| `browser_use.llm.messages`, `.base`, `.views`, `.schema`, `.exceptions` | dropped: the agent calls the model |
| `browser_use.browser.session`, `.profile`, `.views`, `.watchdogs`, `.session_manager`, `.context`, `.cloud`, `ProxySettings` | `from browser_use import Browser` |
| `browser_use.browser.events`, `AgentHookFunc` | a `logging.Handler` on `browser_use` |
| `browser_use.controller`, `browser_use.tools.service`, `.registry`, `.views` | `from browser_use import Tools, ActionResult` |
| `browser_use.agent.views`: `AgentOutput`, `AgentBrain`, `AgentState`, `AgentHistory`, `AgentStepInfo`, `StepMetadata`, `AgentError`, `AgentStructuredOutput` | `AgentHistoryList` and `ActionResult` are kept; the rest is dropped |
| `browser_use.agent.prompts`, `.message_manager`, `.memory`, `.judge`, `.gif`, `.variable_detector`, `SystemPrompt` | `Agent(extend_system_message=...)` |
| `CodeAgent`, `browser_use.code_use` | `Agent(...)`: the 1.0 agent writes Python itself |
| `ActionModel` | `@tools.action(...)` on a plain function |
| `browser_use.dom`, `DomService`, `browser_use.actor` (`Page`, `Element`, `Mouse`) | dropped: the agent drives pages with Python and raw CDP |
| `browser_use.filesystem` | `Agent(workspace=...)` |
| `browser_use.tokens` | `result.usage`, `result.cost_usd` |
| `browser_use.observability` (Laminar `observe`) | Pydantic AI's OpenTelemetry spans; see [Observability](https://docs.browser-use.com/open-source/observability.md) |
| `browser_use.mcp.client`, `browser_use.mcp.controller` | call the MCP server from a plain `@tools.action` function |
| `browser_use.mcp.server`, `browser_use.skill_cli` | `browser-use --cli-mcp` and the `browser-use` CLI |
| `browser_use.config`, `browser_use.utils`, `browser_use.logging_config` | 0.x internals: `Agent(...)` / `Browser(...)` arguments, `BROWSER_USE_LOGGING_LEVEL` |
| `browser_use.beta`, `browser_use.sandbox`, `browser_use.sync`, `browser_use.screenshots`, `browser_use.skills.service`, `browser_use.init_cmd` | dropped (`screenshot()` in the agent's REPL replaces `screenshots`) |
| `browser_use.integrations` (Anthropic and OpenAI hosted-browser tools, Gmail) | not yet: planned on the new `Browser` in a later release. Stay on 0.x for them; see [Claude, OpenAI and Gemini](https://docs.browser-use.com/open-source/integrations.md#0-x-integrations) |

## CLI and MCP

The CLI still runs piped Python, and `browser-use skill install` and `browser-use --cli-mcp` work as before. New: `browser-use login` / `logout` and a `browser_login` MCP tool. `browser-use run`, `init` and the 0.x subcommands (`open`, `state`, `click`, ...) print what replaced them.

The 0.x `browser-use --mcp` server ran the removed engine. `--mcp` now prints a deprecation notice and serves the `--cli-mcp` tools (`browser_exec`, `browser_screenshot`, `browser_login`) on its own temporary Chromium. The old tool names (`browser_navigate`, `browser_click`, `browser_type`, `browser_get_state`, `browser_extract_content`, `browser_get_html`, `browser_scroll`, `browser_go_back`, `browser_list_tabs`, `browser_switch_tab`, `browser_close_tab`, `retry_with_browser_use_agent`, `browser_list_sessions`, `browser_close_session`, `browser_close_all`) return an error that points to `browser_exec`. Switch clients to `--cli-mcp`:

```json theme={"theme":"css-variables"}
{
  "mcpServers": {
    "browser-use": {
      "command": "uvx",
      "args": ["--from", "browser-use==1.0.0", "browser-use", "--cli-mcp"]
    }
  }
}
```

See [Browser Use CLI](https://docs.browser-use.com/open-source/browser-use-cli.md) and [MCP server](https://docs.browser-use.com/open-source/mcp.md).

## Environment variables

| Variable | 1.0 |
| - | - |
| `BROWSER_USE_API_KEY` | read; wins over `browser-use login` |
| `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GOOGLE_API_KEY` | read |
| `ANONYMIZED_TELEMETRY` | read; see [Telemetry](https://docs.browser-use.com/open-source/telemetry.md) |
| `BROWSER_USE_LOGGING_LEVEL` | read |
| `DEFAULT_LLM`, `BROWSER_USE_HEADLESS` | not read, with a warning |
| `BROWSER_USE_ALLOWED_DOMAINS` | not read: `Browser()` raises while it is set |
| `BROWSER_USE_LLM_URL` | not read: `ChatBrowserUse(base_url=...)` |
| `AZURE_OPENAI_KEY` | not read: `AZURE_OPENAI_API_KEY` |

The same guide, with every detail of domain restrictions and the secrets trust model, is [MIGRATING.md](https://github.com/browser-use/browser-use/blob/main/MIGRATING.md) in the repo.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.