Browserbase cookbook

Find a working starting point for browser automation. Browse by task, choose your language or framework, and open the recipe's setup guide and code.

This cookbook brings together Browserbase's public templates, Playbook patterns, and integrations. It includes the source code, with each project's dependencies kept together. Imported examples have been inspected, but have not all been run against live services.

Caution

Demo and reference code only. Browserbase does not claim that any recipe, integration, target, data source, vendor, or workflow in this repository has been vetted, approved, secured, or validated for production use. Independently review the code and obtain all required authorization before running it. You are responsible for compliance, site terms, privacy, security, costs, and outcomes. Use at your own risk.

Browse every recipe · Use the coding-agent skill · Run your first example · Source provenance

Start here

You want to	Start with
Try Search, Fetch, and a cloud browser without an LLM	Python or TypeScript
Keep your existing Playwright code	Python or TypeScript
Extract structured product data with Stagehand	Python or TypeScript
Save a login and handle MFA	Authentication recipes
Pause an agent for a person	Human-in-the-loop app
Add Browserbase to an agent framework	Integrations

Browse by task

Task	Examples you'll find
Getting started	Browserbase SDK, Playwright, Puppeteer, Selenium, and Go
Authentication	Persistent contexts, MFA, and login flows
Browser features	Proxies, caching, session configuration, and CAPTCHA handling
Downloads and documents	File downloads, uploads, PDFs, and document extraction
Extraction and research	Web research, company data, filings, and structured extraction
Forms and transactions	Forms, applications, and checkout workflows
Agents and human handoff	Computer use, tool calling, and human input
Integrations and orchestration	Framework adapters, durable workflows, and platform packages
Testing and observability	Link checks, traces, and evaluation integrations
Commerce and travel	Pricing, products, flights, and hospitality
Business operations	Finance, healthcare, public records, and internal operations

Search locally without installing dependencies or calling an API.

python3 scripts/catalog.py search "persistent login" --language python
python3 scripts/catalog.py search "convex" --limit 3
python3 scripts/catalog.py search "human handoff" --json

Search is a keyword lookup. It ranks matched terms, public access, and match strength before using collection and recipe ID as tie-breakers. Lifecycle does not outrank a stronger task match. The coding-agent skill handles interpreting a task and comparing recipes.

Quick start

From this repository's root, run the Python example's independent browser mode. It creates a cloud session without requiring Search or Fetch access.

cd examples/python/getting-started-with-browserbase
cp .env.example .env

Set BROWSERBASE_API_KEY in .env, then run with uv.

uv run python main.py browser

The script creates a Browserbase session, extracts content from Wikipedia, and prints its replay URL. Use search, fetch, or all instead of browser when you want those separately permissioned APIs. No model API key is required.

For TypeScript setup, expected output, and troubleshooting, read Run your first example.

For coding agents

One skill routes a request to the relevant topic, then to the recipe guide and executable source. It respects your language and framework and identifies relevant Playbook patterns.

Read the skill or install it for Codex, Claude Code, or Cursor.

Example requests include:

• “Find a Python example that reuses a logged-in session.”
• “Add Browserbase to a Convex app.”
• “Build a TypeScript agent that can pause for human input.”
• “Download financial statements and extract their contents.”

Repository layout

examples/       Standalone Python, TypeScript, and Go templates
integrations/   Framework examples and integration packages
playbook/       Browserbase Playbook patterns and shared helpers
skills/         One cookbook skill with focused topic references
docs/           Recipe guides, task indexes, setup, and maintenance
scripts/        Import, catalog generation, search, and validation

Install dependencies inside the selected example. There is no root application or shared dependency installation. The integrations workspace and Playbook Node scripts have their own setup boundaries.

Status and maintenance

current identifies recipes that use the cookbook's supported dependency baseline. It does not promise runtime verification. legacy is reserved for recipes that still require an SDK migration. Read each recipe's manifest and caveats before running it.

Verification records exactly what was checked. Contributing explains how to add or update a recipe. Source provenance records the upstream repositories, pinned commits, import exclusions, and license boundaries.

python3 scripts/verify.py

The catalog and routing experience take inspiration from the Fireworks AI cookbook. The Browserbase code comes from the three repositories recorded in the source manifest.

Around the ecosystem

Track Stagehand on GitHub, the browser automation framework used throughout this cookbook.

Explore Stagehand’s interactive star history.

The live badges show the upstream Stagehand project.

_{Official artwork from the Browserbase media kit. Asset provenance.}