# Browserbase — Cloud browser infrastructure and APIs for building and running browser automation and AI agents

> Source: The Agents Index — https://theagentsindex.com/browserbase (structured, researched, re-verified)
> Facts last verified: 2026-08-17

Browserbase is a cloud browser infrastructure company, founded in 2024 and based in San Francisco, that gives developers programmatic access to headless browsers for automation and AI agents. It ships REST APIs (Sessions, Contexts, Search, Fetch, Agents, and more), Node.js and Python SDKs, and Stagehand, an open-source browser automation framework with 22k GitHub stars. The company has raised $67.5M across three rounds and reports over 100,000 developers and 10,000+ companies using its platform, serving teams that need reliable, scriptable browser access without running their own infrastructure.

| Fact | Value |
| --- | --- |
| Website | https://www.browserbase.com |
| Pricing | Free plan available; paid plans start at $20/mo (Developer), then $99/mo (Startup), with custom Scale pricing |
| API | Yes |
| Best for | Developers building browser agents who want one API instead of managing browser infra |
| Not for | Teams that need the fastest possible cold starts or free anti-detection out of the box: independent benchmarks show Browserbase's session start (5-10s) trailing Kernel and TinyFish, and its free tier ships with no stealth features at all, unlike Kernel's free tier which includes basic stealth and residential proxies. |

## Pricing

| Tier | Price |
| --- | --- |
| Free | $0 / mo |
| Developer | $20 / mo |
| Startup | $99 / mo |
| Scale | Custom |

## Verdict

Browserbase's pitch is a single API key that replaces a homegrown fleet of headless browsers, and the surface area backs that up: Sessions, Contexts, Search, Fetch, and Runtime APIs, automatic recording on every session, and a genuinely popular open-source SDK in Stagehand (22k stars, 700k+ weekly downloads). Where it gives ground is speed and access: independent benchmarks put its cold start around 5-10 seconds, well behind Kernel and TinyFish, and its free tier strips out stealth and anti-detection entirely, pushing anyone scraping protected sites onto a paid plan fast. For teams that want Playwright-level control with less infrastructure to babysit, it's a solid default; for teams chasing the fastest or cheapest scrape, worth comparing against the alternatives first.

## How it works

1. Get an API key and install the SDK or unified CLI (`npm install -g browse`).
2. Create a browser session through the Sessions API, then drive it with Playwright, Puppeteer, Selenium, or Stagehand.
3. Reuse logged-in state across runs with the Contexts API, or call the Search and Fetch APIs directly for lighter-weight data retrieval without a full session.
4. Deploy the agent to run on a schedule or on demand via the Runtime API, with every session auto-recorded for replay and debugging.

## Who it's for

- Developers building AI agents that need to browse and act on real websites
- Teams running web scraping or data extraction at scale without managing their own browser infrastructure
- QA engineers running Playwright, Puppeteer, or Selenium test suites against cloud browsers instead of local ones

## Strengths and weaknesses

- ✓ Broad, unified API surface (Sessions, Contexts, Search, Fetch, Runtime, Agents) covers most browser-automation needs behind one API key.
- ✓ Automatic session recording with video replay, DevTools access, and network/console logs is on by default, unlike Kernel, where recording needs manual activation and is paid-tier only.
- ✓ Stagehand SDK has real community traction: 22k GitHub stars and 700k+ weekly downloads.
- ✓ A genuine free plan exists for testing before committing to a paid tier.
- ✗ Cold start lags newer entrants: about 5-10 seconds per session versus TinyFish's sub-250ms, and roughly 3.4x slower than Kernel's unikernel-based browsers, per independent benchmarks.
- ✗ The free tier ships with no stealth or anti-detection features at all, while Kernel's free tier includes basic stealth and residential proxy support; Browserbase requires at least the Developer plan for basic stealth, and advanced stealth is Scale-only.
- ✗ Search API overage is comparatively pricey at $7 per 1,000 calls, versus TinyFish, which keeps Search and Fetch free on every plan including pay-as-you-go.
- ✗ Still relies on maintaining Playwright, Puppeteer, or Selenium scripts that can break when target sites change layout, unlike vision/LLM-driven tools such as Skyvern.
- ⚠ Overage pricing figures differ slightly between the vendor's own pricing page and third-party comparison sites (e.g., $0.12/hr vs. a $0.10-0.12/hr range), so current numbers should be confirmed on the live pricing page.
- ⚠ HIPAA compliance, DPA, and SSO are only available on the custom-priced Scale plan, not on Free, Developer, or Startup.

## Key features

- **Sessions API** — Create, manage, and observe cloud browser instances programmatically, each a single browser instance running in the cloud.
- **Fetch API** — Converts any URL into HTML, JSON, or clean markdown by routing the request through a proxy network.
- **Search API** — Returns fast, token-efficient web search results so agents can find relevant sites from a single query.
- **Automatic session recording** — Every session is recorded by default with synchronized video replay, Chrome DevTools access, network requests, and console logs.
- **Contexts API** — Configures and reuses browser environments, including logged-in state, across multiple sessions.
- **Stagehand SDK** — Open-source browser automation framework with 22k GitHub stars and 700k+ weekly downloads, built for AI agent workflows.

## Use cases

- **Web scraping and data extraction** — A team uses the Fetch and Search APIs to pull structured page content and search results at scale without maintaining its own proxy or browser pool.
- **AI agent web browsing** — An AI agent uses the Sessions and Runtime APIs to navigate multi-step web workflows, such as filling forms or completing checkout flows, on a schedule or on demand.
- **QA and regression testing** — A QA team runs existing Playwright or Selenium test suites against Browserbase's cloud browsers instead of provisioning and patching local browser infrastructure.
- **Scheduled automation jobs** — A developer deploys a browser agent via the Runtime API to run on a cron, with the Contexts API reusing a logged-in session across each scheduled run.

## Integrations

Playwright · Puppeteer · Selenium · LangChain · CrewAI · Mastra · Claude Code (skill plugin)

## Sources

- https://www.browserbase.com
- https://docs.browserbase.com/reference/introduction
- https://docs.browserbase.com/integrations/skills/browserbase-cli
- https://www.browserbase.com/pricing
- https://www.browserbase.com/plans
- https://www.respan.ai/market-map/browserbase
- https://www.browserbase.com/docs
- https://www.skyvern.com/blog/browserbase-vs-kernel-which-is-better/
- https://www.tinyfish.ai/blog/tinyfish-vs-browserbase