[Blog](https://scrapfly.io/blog) / [ai](https://scrapfly.io/blog/tag/ai) / [How to Create an AI Browser Agent for Free](https://scrapfly.io/blog/posts/how-to-create-an-ai-browser-agent-for-free) # How to Create an AI Browser Agent for Free by [Hisham](https://scrapfly.io/blog/author/hisham) Apr 18, 2026 8 min read [\#ai](https://scrapfly.io/blog/tag/ai) [\#headless-browser](https://scrapfly.io/blog/tag/headless-browser) [\#python](https://scrapfly.io/blog/tag/python) [\#typescript](https://scrapfly.io/blog/tag/typescript) [ ](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fscrapfly.io%2Fblog%2Fposts%2Fhow-to-create-an-ai-browser-agent-for-free "Share on LinkedIn") AI agents that control web browsers through natural language are here, and you can build one today without spending a dollar. Browser-Use and Stagehand are the two most popular open-source frameworks for AI browser automation, with a combined 100k+ GitHub stars between them. Both run locally, use Google's free Gemini API, and give you full control over the source code. Starting free means no vendor lock-in and you learn how these tools actually work before committing to any paid service. In this guide, we'll compare both frameworks head-to-head, then walk through a full quickstart for each with working code examples. Let's get started! [Guide to Understanding and Developing LLM AgentsExplore how LLM agents transform AI, from text generators into dynamic decision-makers with tools like LangChain for automation, analysis & more!](https://scrapfly.io/blog/posts/practical-guide-to-llm-agents) ## Stagehand vs Browser-Use: Pick Your Path Before writing any code, pick the framework that fits your stack and your workflow. Both build on Playwright, so you get the same browser engine underneath. The difference is how they use AI. | Feature | Stagehand | Browser-Use | |---|---|---| | Language | TypeScript, Python, Go, Ruby | Python | | Built on | Playwright | Playwright | | Approach | Hybrid (AI + code) | Pure agent | | Core API | `act()`, `extract()`, `observe()` | `Agent.run()` | | Free LLM | Gemini | Gemini, Ollama | | GitHub Stars | ~21k | ~78k | | Best for | Precise control | Full autonomy | ### When to Choose Browser-Use You're a Python developer who wants full agent autonomy. You give the agent a task in plain English, and it figures out the clicks, scrolling, and extraction on its own. Browser-Use ships with built-in model wrappers for OpenAI, Anthropic, and Google Gemini, so setup is a single import. More importantly, Browser-Use supports [Ollama](https://ollama.com/) for fully local inference. Ollama is the only option that costs literally zero dollars. No API key, no rate limits, no data leaving your machine. If privacy or offline use matters, Browser-Use with Ollama is the way to go. ### When to Choose Stagehand You want a hybrid approach that mixes AI with traditional code. Stagehand gives you three AI-powered primitives: `act()` for clicking and typing, `extract()` for pulling structured data, and `observe()` for understanding page state. You combine these with standard Playwright commands and keep full control over the execution flow. This makes scripts easier to debug and maintain. When something breaks, you know exactly which step failed because you wrote the flow yourself. Browserbase maintains official Stagehand SDKs in TypeScript, Python, Go, and Ruby. TypeScript is the primary SDK with the most features, but the Python SDK mirrors the same API using Pydantic models for schema validation. [What is a Headless Browser? Top 5 Headless Browser ToolsQuick overview of new emerging tech of browser automation - what exactly are these tools and how are they used in web scraping?](https://scrapfly.io/blog/posts/what-is-a-headless-browser-top-5-headless-browser-tools) ## Quickstart: Browser-Use (Python) Browser-Use takes a "pure agent" approach. You describe a task in natural language, and the agent handles navigation, clicking, scrolling, and data extraction on its own. Let's build one. ### Prerequisites - Python 3.11+ - A Google account for a free [Gemini API key](https://aistudio.google.com) - Playwright browsers installed locally Get your free Gemini API key at [Google AI Studio](https://aistudio.google.com). Click "Get API key" in the left down sidebar, then click "Create API key." No credit card needed. The free tier gives you 10 requests per minute with Gemini 2.5 Flash, which is plenty for step-by-step agent development. ### Building Your First Agent #### Installation shell```shell pip install browser-use playwright install chromium ``` #### Agent Code python```python import os import asyncio from browser_use import Agent, ChatGoogle os.environ["GOOGLE_API_KEY"] = "Your Google API Key" async def main(): llm = ChatGoogle(model="gemini-flash-latest") agent = Agent( task=( "Go to https://web-scraping.dev/products and extract " "the name and price of every product on this page only" ), llm=llm, ) result = await agent.run() print(result.final_result()) if __name__ == "__main__": asyncio.run(main()) ``` Browser-Use ships with a built-in `ChatGoogle` wrapper, so you don't need any extra LangChain packages. The `Agent` class takes a natural language task and an LLM, launches a browser, and extracts data without a single CSS selector. Browser-Use also works with [Ollama](https://ollama.com/) for fully local inference. Swap `ChatGoogle` for a local model and you won't need any API key at all. Check the Browser-Use docs for supported model setup. If you need typed output instead of plain text, Browser-Use supports Pydantic models for structured extraction. The official documentation covers this in the output models section. [Web Scraping with Playwright and PythonPlaywright is the new, big browser automation toolkit - can it be used for web scraping? In this introduction article, we'll take a look how can we use Playwright and Python to scrape dynamic websites.](https://scrapfly.io/blog/posts/web-scraping-with-playwright-and-python) ## Quickstart: Stagehand (TypeScript) Stagehand takes a different path. Instead of handing everything to an autonomous agent, it gives you three AI-powered methods that you weave into regular Playwright code. You control the flow. AI handles the parts where writing selectors would be fragile or tedious. ### Prerequisites - Node.js 18+ - A Google account for a free [Gemini API key](https://aistudio.google.com) You'll use the same free Gemini API key from Google AI Studio. If you already got one for the Browser-Use example above, it will work here too. ### Building Your First Agent #### Installation Scaffold a new project with the Stagehand CLI: shell```shell npx create-browser-app my-agent cd my-agent && npm install ``` Or add Stagehand to an existing project: shell```shell npm i @browserbasehq/stagehand zod dotenv ``` #### Agent Code typescript```typescript process.env.GOOGLE_GENERATIVE_AI_API_KEY = "Your Google API Key"; import { Stagehand } from "@browserbasehq/stagehand"; import { z } from "zod"; async function main() { const stagehand = new Stagehand({ env: "LOCAL", model: "google/gemini-2.5-flash", }); await stagehand.init(); const page = stagehand.context.pages()[0]; await page.goto("https://web-scraping.dev/products"); await stagehand.act("Click on the first product link"); const product = await stagehand.extract( "Extract the product details from this page", z.object({ name: z.string().describe("the product name"), price: z.string().describe("the product price"), description: z.string().describe("the product description"), }), ); console.log("Product:", product); await stagehand.close(); } main(); ``` You write the navigation flow with `page.goto()` and control the order of operations. AI steps in only where you call `stagehand.act()` or `stagehand.extract()`. If the page data doesn't match your Zod schema, you get a validation error right away. The `observe()` method helps when you're exploring an unfamiliar page. Call `await stagehand.observe("What links are available?")` and Stagehand returns structured observations about the current page state. It's useful for building scripts that need to adapt to changing page layouts. Stagehand also supports a pure agent mode for fully autonomous tasks, similar to Browser-Use. See the Stagehand documentation for the agent API and advanced patterns. [Playwright vs SeleniumExplore the key differences between Playwright vs Selenium in terms of performance, web scraping, and automation testing for modern web applications.](https://scrapfly.io/blog/posts/playwright-vs-selenium) Scrapfly #### Need a cloud browser for scraping? Run headless browsers at scale with Scrapfly Cloud Browser — no infrastructure to manage. [Try Free →](https://scrapfly.io/register)## When Free Isn't Enough Running agents on your local machine works well for development and small projects. But local setups hit real limits once you try to scale up or target protected sites. A single machine can only run one browser at a time without extra setup. Anti-bot systems flag your home or server IP quickly. Modern detection systems check browser fingerprints like canvas rendering, WebGL hashes, font lists, and TLS signatures. They spot automated browsers almost right away. On top of that, managing sessions, cookies, and browser state across long-running tasks gets messy fast. When you're ready for production workloads, need to get past anti-bot systems, or want to run multiple agents in parallel, cloud browsers are the practical next step. [Scrapfly's Cloud Browser](https://scrapfly.io/docs/cloud-browser-api/getting-started) works with both Browser-Use and Stagehand. Your agent code stays the same. You just swap the local browser connection for a cloud endpoint. Built-in residential proxies and managed fingerprinting let your agents access sites that would block a local browser right away. For step-by-step setup, check the [Stagehand + Scrapfly integration guide](https://scrapfly.io/docs/cloud-browser-api/stagehand) and the [Browser-Use + Scrapfly integration guide](https://scrapfly.io/docs/cloud-browser-api/browser-use): typescript```typescript import { Stagehand } from "@browserbasehq/stagehand"; async function main() { // initialize Stagehand with external CDP browser connection const stagehand = new Stagehand({ env: "LOCAL", // local means are we aren't using browserbase, we are using a Scrapfly's cloud browser over cdp localBrowserLaunchOptions: { // connect to Scrapfly's cloud browser via CDP cdpUrl: "wss://browser.scrapfly.io?key=YOUR_API_KEY" } }); // initialize the stagehand await stagehand.init(); // create an agent with specific model configuration const agent = stagehand.agent({ model: { // choose the llm model to use and add its api key modelName: "openai/computer-use-preview", apiKey: "YOUR_API_KEY" }, // add a system prompt to the agent systemPrompt: "You are an AI browser automation agent. Follow instructions precisely." }); // define the task to be performed by the agent const task = ` Go to https://web-scraping.dev/products Extract the product names and prices Return the data as JSON `; // execute the agent workflow const result = await agent.execute(task); console.log("Agent workflow result:", result); // close the connection await stagehand.close(); } main(); ``` [Best Cloud Browser APIs in 2026Comprehensive comparison of cloud browser APIs for developers. Evaluate Bright Data, Browserless, Browserbase, Scrapfly, Steel.dev, and more with transparent pricing, feature matrices, and code examples.](https://scrapfly.io/blog/posts/best-cloud-browser-apis) ## FAQ Is Browser-Use really free?Yes. Browser-Use is open-source and works with Google Gemini's free tier or fully local models through Ollama. Can I use Stagehand with Python?Yes. Browserbase maintains an official Python SDK that mirrors the TypeScript API using Pydantic models. Do I need a paid API key for either framework?No. Both work with Google Gemini's free tier, which gives you 10 requests per minute at no cost. ## Summary You've built two free AI browser agents from scratch. Browser-Use gives you a Python-first, fully autonomous agent that figures out navigation and extraction from a single task description. Stagehand gives you fine-grained control with AI-powered primitives you can mix into Playwright scripts. Both frameworks are actively maintained and ready for real work. Pick Browser-Use if you want pure autonomy in Python. Go with Stagehand if you want hybrid control in TypeScript. Either way, you're running a real AI browser agent without spending a cent. Try the examples above with your free Gemini key, then explore [Scrapfly's Cloud Browser](https://scrapfly.io/docs/cloud-browser-api/getting-started) when you're ready to scale past your local machine. Table of Contents Table of Contents- [Stagehand vs Browser-Use: Pick Your Path](#stagehand-vs-browser-use-pick-your-path) - [When to Choose Browser-Use](#when-to-choose-browser-use) - [When to Choose Stagehand](#when-to-choose-stagehand) - [Quickstart: Browser-Use (Python)](#quickstart-browser-use-python) - [Prerequisites](#prerequisites) - [Building Your First Agent](#building-your-first-agent) - [Quickstart: Stagehand (TypeScript)](#quickstart-stagehand-typescript) - [Prerequisites](#prerequisites) - [Building Your First Agent](#building-your-first-agent) - [When Free Isn't Enough](#when-free-isn-t-enough) - [FAQ](#faq) - [Summary](#summary) Join the Newsletter Get monthly web scraping insights Scale Your Web Scraping Anti-bot bypass, browser rendering, and rotating proxies, all in one API. Start with 1,000 free credits. No credit card required 1,000 free API credits Anti-bot bypass included [Start Free](https://scrapfly.io/register) [View Docs](https://scrapfly.io/docs/onboarding) Not ready? Get our newsletter instead. ## Explore this Article with AI [ ChatGPT ](https://chat.openai.com/?q=Summarize%20this%20page%3A%20https%3A%2F%2Fscrapfly.io%2Fblog%2Fposts%2Fhow-to-create-an-ai-browser-agent-for-free) [ Gemini ](https://www.google.com/search?udm=50&aep=11&q=Summarize%20this%20page%3A%20https%3A%2F%2Fscrapfly.io%2Fblog%2Fposts%2Fhow-to-create-an-ai-browser-agent-for-free) [ Grok ](https://x.com/i/grok?text=Summarize%20this%20page%3A%20https%3A%2F%2Fscrapfly.io%2Fblog%2Fposts%2Fhow-to-create-an-ai-browser-agent-for-free) [ Perplexity ](https://www.perplexity.ai/search/new?q=Summarize%20this%20page%3A%20https%3A%2F%2Fscrapfly.io%2Fblog%2Fposts%2Fhow-to-create-an-ai-browser-agent-for-free) [ Claude ](https://claude.ai/new?q=Summarize%20this%20page%3A%20https%3A%2F%2Fscrapfly.io%2Fblog%2Fposts%2Fhow-to-create-an-ai-browser-agent-for-free) ## Related Articles [ typescript scrapeguide ### Complete Guide to Web Scraping using Typescript In this introduction we're taking a look at web scraping using Typescript - increasingly popular typed Javascript langua... ](https://scrapfly.io/blog/posts/ultimate-intro-to-web-scraping-with-typescript) [ python playwright ### How to Scrape Google Maps We'll take a look at to find businesses through Google Maps search system and how to scrape their details using either S... ](https://scrapfly.io/blog/posts/how-to-scrape-google-maps) [ python scrapeguide ### How to Scrape Google Search Results in 2026 In this scrape guide we'll be taking a look at how to scrape Google Search - the biggest index of public web. We'll cov... ](https://scrapfly.io/blog/posts/how-to-scrape-google) Run headless browsers at scale, **1,000 free credits** [Start Free](https://scrapfly.io/register)