← Back to stack

Builder infrastructure

Browserbase

Give your agents access to the whole web

Visit official site →

Browserbase is a cloud platform that gives AI agents real, programmable browsers so they can interact with websites the way a human would - handling logins, dynamic content, CAPTCHAs, and multi-step workflows - rather than being limited to the subset of the web that exposes an API.

Vendor's own claim

What Browserbase is

Vendor's own claim

Browserbase is browser infrastructure for AI agents: it runs real Chromium sessions in the cloud so agents can log in, click, fill forms, and scrape content on sites without a public API. It's used both directly via a REST API/SDKs and through a hosted MCP server, and it underlies the open-source Stagehand automation framework.

What it can do

Vendor's own claim
  • Browsers: on-demand, scalable headless browser sessions with full programmatic control
  • Search API: query-based web search built for agent consumption
  • Runtime: sandboxed, scalable deployment environments for agent code
  • Identity: proxying and agent authentication for human-like web navigation, plus built-in CAPTCHA solving
  • Agents: fully managed browser agents driven by a prompt
  • Model Gateway: single API key for routing to multiple LLM providers
  • Observability: session replays, logs, and prompt-level debugging
  • Stagehand: open-source browser automation framework built by Browserbase
  • Hosted MCP server for LLM-driven browser automation (start, navigate, act, observe, extract)

Who it is for

Vendor's own claim

Teams building AI agents or automation products that need agents to operate on arbitrary websites - covering logins, dynamic JS-heavy pages, and CAPTCHAs - rather than only on sites with a public API. Vendor case studies name Stripe, Vercel, Okta, Anthropic, Clay, Amplitude, and Ramp as customers.

When to choose something else

Our read

The Free plan gives only 1 browser-hour and 3 concurrent sessions with a 7-day data retention window, so it only works for prototyping, not production load.

Implementation considerations

Our read

Setup assessment: 30m

Creating an account and running a first session via the Node.js or Python SDK is a sub-hour task once an API key is issued. Production setup - proxy configuration, CAPTCHA handling, MCP server wiring, or Model Gateway routing - takes longer and is not a 5-minute quickstart.

Prerequisites

  • Browserbase account and API key
  • For MCP: an MCP-compatible client configured with the hosted MCP URL and API key as a query parameter

Developer and agent access

From official docs
Authentication
API key passed in the "x-bb-api-key" request header
Base URL
https://api.browserbase.com/v1/
Rate limits
Concurrency and session-creation limits vary by plan (e.g.
Machine-readable API
OpenAPI specification linked
  • Sessions API - create/manage/retrieve browser sessions (/v1/sessions/{sessionId})
  • Projects API - project-wide usage metrics
  • Contexts API - configure and reuse browser environments across sessions

Published pricing

From official docs
PlanPriceIncluded
Free$0/moSee official pricing
Developer$20/moSee official pricing
Startup$99/moSee official pricing
ScaleCustomSee official pricing

Verify before you adopt it

  • Exact GitHub org-level repo count or stars for repos other than mcp-server-browserbase
  • Whether the Free plan has a hard expiration date beyond the quotas listed
  • LinkedIn/X/YouTube follower counts