Structured output

Get validated, typed data back from agent tasks.

Pass a Pydantic model (Python) or Zod schema (TypeScript) — result.output is automatically validated and converted to the typed object.

TypeScript requires Zod v4 (npm install zod@4). Zod v3 is not compatible.

Python

from browser_use_sdk.v3 import AsyncBrowserUse
from pydantic import BaseModel

class Post(BaseModel):
    name: str
    points: int
    comments: int

class HNPosts(BaseModel):
    posts: list[Post]

client = AsyncBrowserUse()
result = await client.run(
    "List the top 20 posts on Hacker News today with their points",
    output_schema=HNPosts,
)
for post in result.output.posts:
    print(f"{post.name} ({post.points} pts, {post.comments} comments)")

Scrape multiple sites concurrently

Each client.run() call creates its own session, so you can run them in parallel with asyncio.gather (Python) or Promise.all (TypeScript). Every result is fully typed.

Python

import asyncio
from browser_use_sdk.v3 import AsyncBrowserUse
from pydantic import BaseModel

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

class Results(BaseModel):
    products: list[Product]

async def main():
    client = AsyncBrowserUse()

    sites = [
        "Find the top 3 laptops on amazon.com with their prices",
        "Find the top 3 laptops on bestbuy.com with their prices",
        "Find the top 3 laptops on newegg.com with their prices",
    ]

    results = await asyncio.gather(*[
        client.run(task, output_schema=Results)
        for task in sites
    ])

    for result in results:
        for p in result.output.products:
            print(f"{p.name} — {p.price}")

asyncio.run(main())