browser-use-sdk 3.0.3

Python SDK for the Browser Use cloud API

browser-use-sdk

Official Python SDK for the Browser Use cloud API.

Run AI-powered browser agents, manage sessions, profiles, skills, and more -- all from Python with both sync and async interfaces. Fully typed with Pydantic models.

Installation

pip install browser-use-sdk

Or with your preferred package manager:

# uv
uv add browser-use-sdk

# poetry
poetry add browser-use-sdk

Quick Start

from browser_use_sdk import AsyncBrowserUse

client = AsyncBrowserUse()
result = await client.run("Go to google.com and search for 'browser use'")
print(result.output)
print(result.id, result.status)

v2 API (default)

The default import gives you the v2 client with access to all resources:

from browser_use_sdk import AsyncBrowserUse, BrowserUseError

client = AsyncBrowserUse()

result = await client.run("Navigate to example.com and get the title")

# Or use the tasks resource directly
task = await client.tasks.create("Navigate to example.com and get the title")
result = await client.tasks.get(str(task.id))
await client.tasks.stop(str(task.id))

# Sessions
session = await client.sessions.create()
sessions = await client.sessions.list(page_size=20)
await client.sessions.stop(str(session.id))
await client.sessions.delete(str(session.id))

# Profiles
profile = await client.profiles.create(name="my-profile")
await client.profiles.update(str(profile.id), name="renamed")
await client.profiles.delete(str(profile.id))

# Browsers
browser = await client.browsers.create()
await client.browsers.stop(str(browser.id))

# Billing
account = await client.billing.account()
print(account.total_credits_balance_usd)

# Skills
skills = await client.skills.list()
await client.skills.execute(skill_id, input="some input")

# Marketplace
marketplace_skills = await client.marketplace.list()

# Files
url_info = await client.files.session_url(session_id, file_name="doc.pdf", content_type="application/pdf", size_bytes=1024)
output = await client.files.task_output(task_id, file_id)

Structured Output

Pass a Pydantic model to get typed results:

from pydantic import BaseModel
from browser_use_sdk import AsyncBrowserUse

class Product(BaseModel):
    name: str
    price: float

client = AsyncBrowserUse()
result = await client.run("Find the price of the MacBook Air", schema=Product)
print(f"{result.output.name}: ${result.output.price}")

Streaming

client.run() returns a dual-purpose handle: await it for the output, or async for it for step-by-step progress.

from browser_use_sdk import AsyncBrowserUse

client = AsyncBrowserUse()
async for step in client.run("Scrape the front page of HN"):
    print(f"[{step.number}] {step.next_goal} — {step.url}")

v3 API (experimental)

The v3 API uses a simplified session-based model:

from browser_use_sdk.v3 import AsyncBrowserUse

client = AsyncBrowserUse()
output = await client.run("Search for Browser Use on Google")
print(output)

# Or use sessions resource directly
session = await client.sessions.create("Search for Browser Use on Google")
result = await client.sessions.get(str(session.id))
files = await client.sessions.files(str(session.id))
await client.sessions.stop(str(session.id))

Sync Usage

A synchronous client is also available:

from browser_use_sdk import BrowserUse

client = BrowserUse()
result = client.run("Go to example.com")

# Streaming (sync)
for step in client.stream("Scrape the front page of HN"):
    print(f"[{step.number}] {step.next_goal} — {step.url}")

Error Handling

All non-2xx responses raise BrowserUseError:

from browser_use_sdk import AsyncBrowserUse, BrowserUseError

client = AsyncBrowserUse()
try:
    await client.tasks.get("nonexistent-id")
except BrowserUseError as e:
    print(e.status_code)  # 404
    print(e.message)      # "Task not found"
    print(e.detail)       # Full response body (dict or None)

Retries are automatic: the SDK retries up to 3 times with exponential backoff on 429 (rate limit) responses.

Resource Methods

v2

Resource	Method	HTTP
billing	account()	GET /billing/account
tasks	create(task, **kw)	POST /tasks
tasks	list(**kw)	GET /tasks
tasks	get(id)	GET /tasks/{id}
tasks	stop(id)	PATCH /tasks/{id}
tasks	status(id)	GET /tasks/{id}/status
tasks	wait(id, **kw)	Poll until terminal, return TaskView
tasks	logs(id)	GET /tasks/{id}/logs
sessions	create(**kw)	POST /sessions
sessions	list(**kw)	GET /sessions
sessions	get(id)	GET /sessions/{id}
sessions	stop(id)	PATCH /sessions/{id}
sessions	delete(id)	DELETE /sessions/{id}
sessions	get_share(id)	GET /sessions/{id}/public-share
sessions	create_share(id)	POST /sessions/{id}/public-share
sessions	delete_share(id)	DELETE /sessions/{id}/public-share
files	session_url(id)	POST /files/sessions/{id}/presigned-url
files	browser_url(id)	POST /files/browsers/{id}/presigned-url
files	task_output(t, f)	GET /files/tasks/{t}/output-files/{f}
profiles	create(**kw)	POST /profiles
profiles	list(**kw)	GET /profiles
profiles	get(id)	GET /profiles/{id}
profiles	update(id, **kw)	PATCH /profiles/{id}
profiles	delete(id)	DELETE /profiles/{id}
browsers	create(**kw)	POST /browsers
browsers	list(**kw)	GET /browsers
browsers	get(id)	GET /browsers/{id}
browsers	stop(id)	PATCH /browsers/{id}
skills	create(**kw)	POST /skills
skills	list(**kw)	GET /skills
skills	get(id)	GET /skills/{id}
skills	update(id, **kw)	PATCH /skills/{id}
skills	delete(id)	DELETE /skills/{id}
skills	cancel(id)	POST /skills/{id}/cancel
skills	execute(id, **kw)	POST /skills/{id}/execute
skills	refine(id, **kw)	POST /skills/{id}/refine
skills	rollback(id)	POST /skills/{id}/rollback
skills	executions(id)	GET /skills/{id}/executions
skills	execution_output()	GET /skills/{id}/executions/{eid}/output
marketplace	list(**kw)	GET /marketplace/skills
marketplace	get(slug)	GET /marketplace/skills/{slug}
marketplace	clone(id)	POST /marketplace/skills/{id}/clone
marketplace	execute(id, **kw)	POST /marketplace/skills/{id}/execute

v3

Resource	Method	HTTP
sessions	create(task, **kw)	POST /sessions
sessions	list(**kw)	GET /sessions
sessions	get(id)	GET /sessions/{id}
sessions	stop(id)	POST /sessions/{id}/stop
sessions	files(id, **kw)	GET /sessions/{id}/files

Configuration

client = BrowserUse(
    api_key="bu_...",
    base_url="https://custom-endpoint.example.com/api/v2",  # override base URL
    timeout=60.0,  # request timeout in seconds (default: 30)
)

License

MIT