# Browserbase

> Managed headless-browser infrastructure for AI agents and web automation — serverless cloud browsers with stealth, proxies, live view, and Playwright/Stagehand.

Browserbase is managed headless-browser infrastructure for AI agents and web automation: serverless cloud browsers with stealth, proxies, CAPTCHA handling, and live session view. You drive them with Playwright, Puppeteer, or Browserbase's open-source Stagehand framework. It's proprietary SaaS with a free tier and paid plans; confirm current tiers on the official site.

Website: https://www.browserbase.com

Browserbase is **managed headless-browser infrastructure for AI agents and web automation at scale**. Instead of provisioning, patching, and babysitting your own Chromium fleet, you request a browser from Browserbase's API and get a real, sandboxed cloud session — with stealth, proxies, CAPTCHA handling, and a live view for watching what the agent does. It's aimed at teams running browser agents, scrapers, and automated workflows in production, where reliability and concurrency matter more than a one-off script.

It fits an AI-assisted workflow as the **execution layer**: your agent (or LLM-driven script) decides what to do, and Browserbase supplies the browser it acts in. A useful distinction — **Browserbase is the hosted browsers; [Stagehand](/tools/stagehand) is the open-source framework** (MIT, from the same company) that adds natural-language control and can run either locally or on Browserbase's cloud.

## Highlights

- **Serverless cloud browsers** — request a real browser session over an API and let it scale to many concurrent sessions; no infrastructure to provision or keep patched.
- **Stealth, proxies, and CAPTCHA handling** — built-in residential/datacenter proxies and anti-bot measures help automations reach sites that block naive headless traffic.
- **Live session view and debugging** — watch sessions in real time and review replays, logs, and network activity to diagnose why a run failed.
- **Works with standard SDKs** — drive sessions with Playwright or Puppeteer over CDP, or with Browserbase's own Stagehand framework, in TypeScript or Python.

## In an AI-assisted workflow

Point an existing Playwright script at a Browserbase session instead of a local browser — the automation code barely changes, but the browser now runs in the cloud with stealth and proxies. See [Browser Agents Compared (2026)](/guides/comparisons/browser-agents-compared-2026) for how the hosted-infra approach compares with all-in-one agents.

```ts
import { Browserbase } from "@browserbasehq/sdk";
import { chromium } from "playwright-core";

const bb = new Browserbase({ apiKey: process.env.BROWSERBASE_API_KEY });
const session = await bb.sessions.create({ projectId: process.env.BB_PROJECT_ID });

const browser = await chromium.connectOverCDP(session.connectUrl);
const page = (await browser.contexts())[0].pages()[0];
await page.goto("https://example.com");
// ...drive the page with your agent, then close the session
```

> [!TIP]
> Use the live session view (and replays) early. Most browser-agent failures are a misread page or a blocked request, and watching the actual session is faster than guessing from logs.

## Good to know

Browserbase is a **proprietary SaaS platform**, not open source — the open-source piece is Stagehand, not the browser infrastructure itself. There is a free tier for prototyping plus paid plans (Free, Developer, Startup, and a custom Scale tier at publication), metered on browser-hours, concurrency, and proxy bandwidth; these change frequently, so confirm current tiers and limits on the official pricing page before you budget. For background on the agents that drive these browsers, see [How Computer-Use Agents Work](/guides/concepts/how-computer-use-agents-work).

---

_Source: https://agentscamp.com/tools/browserbase — Tool on AgentsCamp._