# BrowserAct - Describe what you need. Get a reusable scraper for any website.

> BrowserAct turns a prompt describing the data you need into a reusable web scraper. The Agent builds and tests the scraper on the live website, returns structured data, and keeps the Bot ready to run again. When websites change, the Bot can adapt and evolve.

## What is BrowserAct?

BrowserAct is an AI-powered web scraping platform for extracting reliable, structured data from websites accessible through a browser. It combines a prompt-driven Cloud Runtime with an AI Agents CLI. Teams can build reusable scraper Bots without writing code or maintaining selectors, while local Agents can use the same browser automation infrastructure to work across the live web.

The core promise is simple: **From one prompt to a working scraper. Build once, run repeatedly, and refine as websites evolve.**

### Core capabilities

- No-code scraping and automation: Describe the data you need in a prompt and get ready-to-use results.
- AI-powered reliability: Adapt to website changes during runs and refine the Bot through prompts as requirements evolve.
- Ready-to-use data: Output structured records, files, downloads, and run logs.
- Access to the real web: Handle dynamic pages, authenticated sessions, and common CAPTCHA challenges.
- Human assistance: Request remote user input when a workflow requires a person to complete a step.
- Two browser modes: Choose Private or Standard mode based on your business needs.
- High-quality residential proxies: Use built-in static and rotating residential proxies for reliable regional access.
- Multiple ways to use BrowserAct: Run in the cloud or connect through the AI Agents CLI, API, workflows, or MCP.

## BrowserAct capabilities

BrowserAct covers the complete workflow from scraper creation and real-browser execution to repeatable runs, continuous refinement, and result delivery. Users can build reusable Bots from prompts in the cloud or provide the same browser capabilities to local Agents through the AI Agents CLI.

### Supported website types

BrowserAct can collect data from almost any website accessible through a browser, including:

- **E-commerce websites**: Amazon, eBay, Shopify stores, product catalogs, and more.
- **Social media**: LinkedIn profiles, X feeds, Instagram posts, and more.
- **Real estate platforms**: MLS listings, property directories, rental websites, and more.
- **Job websites**: Indeed, LinkedIn Jobs, company career pages, and more.
- **Business directories**: Yellow Pages, Google Maps, Yelp business listings, and more.
- **News and media**: News articles, press releases, blogs, and more.
- **Financial data websites**: Stock prices, cryptocurrency data, market data, and more.
- **Government websites**: Permits, licenses, public records, and more.

### Prompt-driven scraper creation

#### From data requirements to a working Bot

Describe the target website, required fields, filters, number of records, and output format in natural language. The BrowserAct Agent opens the live website, analyzes its structure and interactions, builds and tests the scraper, and saves the working process as a reusable Bot.

No scraper code is required, and there are no CSS or XPath selectors to maintain manually.

#### Supported data collection tasks

- **List and detail extraction**: Products, business directories, jobs, properties, creators, and search results.
- **Multi-page extraction**: Pagination, infinite scroll, load-more flows, and list-to-detail navigation.
- **Search and filtering**: Enter keywords, regions, and business criteria before collecting matching results.
- **Authenticated data**: Reuse an authorized login session to access account dashboards or member-only pages.
- **Files and downloads**: Trigger exports, download files, and deliver them alongside run records.

### Real browser automation

BrowserAct runs tasks in real browsers rather than relying only on static HTML. This allows it to work with JavaScript-rendered pages, asynchronous content, and websites that require interaction.

#### Browser actions

- Open websites and navigate across pages or tabs.
- Click buttons, menus, links, and page controls.
- Enter text, complete forms, and submit search criteria.
- Wait for dynamic content, scroll pages, and handle pagination.
- Upload files, trigger downloads, and retrieve exported results.
- Read live page state and extract structured data.

### Reliable execution and continuous improvement

#### Reusable and repeatable

Once built, a Bot can run repeatedly with different parameters for recurring collection, monitoring, and business data updates. Each run follows the same task objective and structured output requirements, reducing the cost of rebuilding the workflow.

#### Adaptation to website changes

When page elements or interaction paths change during a run, the Agent evaluates the current page state, determines the next action, and attempts to continue the task instead of relying entirely on fixed scripts, selectors, or predefined paths.

#### Continuous refinement through prompts

When website flows, data fields, or business requirements change significantly, describe the new requirements in natural language, rebuild and test the Bot in draft, inspect the results, and publish the updated version. BrowserAct brings website changes and requirement changes into the same iterative workflow.

### Data output and run inspection

#### Output formats

- Structured data and field-based records.
- Ready-to-use data files such as CSV.
- Downloads and exported files generated by the website.
- Run logs and execution status for every task.

#### Delivery and integrations

Results can be used in spreadsheets, analytics, internal applications, or downstream Agent reasoning. They can also be connected to existing workflows through APIs, webhooks, MCP, Make, n8n, and Zapier.

Run history and logs help teams inspect Bot actions, locate failed steps, and identify opportunities for the next iteration.

### Overcoming web access barriers

BrowserAct includes the access infrastructure required for real-world browser automation, helping tasks work with dynamic websites, authenticated sessions, regional restrictions, and common verification challenges.

#### Stealth browsers and residential proxies

- Stealth browser fingerprints and TLS rotation reduce common automation signals.
- Built-in high-quality rotating and static residential proxies support region-specific access.
- More than 30 fingerprint attributes help browser behavior more closely resemble a real user session.

#### Authentication and verification

- Standard Browser mode preserves cookies, login sessions, and core browser profile data.
- Common CAPTCHA challenges are supported, including reCAPTCHA, Cloudflare Turnstile, DataDome, and HUMAN Security.
- When 2FA, QR codes, SMS verification, SSO, or sensitive confirmations require user action, Remote Assist can request human input. After the user completes the step, the Agent continues from the same browser state.

These capabilities improve continuity across real-world web tasks, but no automation system can guarantee access to every website or completion of every verification flow.

### Browser modes and session identity

BrowserAct provides two browser modes for one-time extraction, repeatable workflows, and multi-account operations:

- **Standard Browser**: Reuses a persistent browser profile and retains cookies and browser data. It is designed for authenticated workflows, repeatable runs, and long-lived identities.
- **Private Browser**: Provides an isolated browser for a single build or run and does not retain browsing data after the task. It is designed for one-time extraction and workflows that require environment isolation.

Browser identities can be combined with proxies and authenticated accounts to maintain independent session environments for different tasks or accounts.

### Browser capabilities for AI Agents CLI

The AI Agents CLI brings BrowserAct's browser runtime to Claude Code, Cursor, Codex, OpenClaw, and other local Agents. Agents can work on the live web without implementing browser drivers, page parsing, or anti-automation infrastructure themselves.

#### Agent-native runtime interface

- **State**: Returns a clean, low-token, indexed representation of the current page so the Agent can understand it.
- **Actions**: Provides browser actions such as `navigate`, `click`, `type`, `wait`, and `upload`.
- **Memory**: Uses semantic browser memory to identify and reuse the appropriate browser identity and session.
- **Safety**: Pauses before login, sensitive settings, or steps that require human confirmation and asks the user to decide.

The CLI can reuse authenticated sessions from local Chrome or connect to isolated BrowserAct browsers, residential proxies, and Remote Assist.

## Ways to use BrowserAct

### BrowserAct Cloud Runtime

Start with a prompt and run the task in a managed cloud environment. No local installation or computer storage is required. Start, manage, and repeat scraper tasks from anywhere, at any time.

### BrowserAct AI Agents CLI

Install BrowserAct into a local coding Agent when it needs to work directly on the live web. The CLI is designed for Claude Code, Cursor, OpenClaw, Hermes, Codex, and other compatible Agents or MCP clients. It can reuse local Chrome sessions and provide interactive browser control directly inside the Agent workflow.

### API and workflow execution

Use the BrowserAct API and workflow endpoints when browser tasks need to be triggered by applications, scheduled processes, or internal services. Published Bots can connect to automation platforms such as Make, n8n, and Zapier.

### MCP and Agent integration

Connect BrowserAct to an MCP-compatible AI client, or install open-source BrowserAct skills from the [browser-act/skills repository](https://github.com/browser-act/skills). Existing Agent workflows can use BrowserAct browser capabilities without replacing the rest of their technology stack.

### Choosing an entry point

- Choose **Cloud Runtime** for managed, repeatable scraper tasks that require no local configuration and can run from anywhere, at any time.
- Choose **AI Agents CLI** for local Agent control, local Chrome sessions, and interactive browser tasks.
- Choose **API or workflows** for application integrations, scheduled execution, and automation platforms.
- Choose **MCP** to provide browser capabilities inside compatible Agents or chat clients.

## Pricing

Visit the latest [BrowserAct pricing page](https://www.browseract.com/pricing) for current plans, promotions, trial eligibility, and accurate pricing. Subscription and infrastructure prices may change, and the pricing page is the source of truth for current amounts.

### Subscription plans

BrowserAct currently offers the following subscription plans with monthly and annual billing. Annual plans are 20% off:

| Plan | Price | Credits | Key allowances |
| --- | --- | --- | --- |
| **Free Trial** | $0 | 2,000 Free Credits on signup | Free features, community support, and 5 free Local Browsers. |
| **Basic** | <nobr>$16 / month</nobr><br><nobr>$13 / month annually ($156 / year)</nobr> | 10,000 Credits / month | 10 concurrent tasks, up to 5 Static Proxies, and 10 free Local Browsers. |
| **Essential** | <nobr>$70 / month</nobr><br><nobr>$56 / month annually ($672 / year)</nobr> | 50,000 Credits / month | 20 concurrent tasks, up to 25 Static Proxies, and 50 free Local Browsers. |
| **Advanced** | <nobr>$120 / month</nobr><br><nobr>$96 / month annually ($1,152 / year)</nobr> | 100,000 Credits / month | 40 concurrent tasks, up to 50 Static Proxies, and 100 free Local Browsers. |

All paid plans support additional Credit purchases. Plan prices, promotions, and allowances may change. Refer to the live [BrowserAct pricing page](https://www.browseract.com/pricing) for the latest information.

### Credit-based browser infrastructure

The current pricing page lists the following usage rates:

- **Workflow Step**: 5 Credits per step, starting from approximately $0.0032 per step.
- **Local Browser**: 100 Credits per browser, starting from approximately $0.064 per browser.
- **Rotating Proxy**: 5,000 Credits per GB, starting from approximately $3.20 per GB.
- **Static Proxy**: Billed per proxy per month; pricing varies by region and plan tier.
- **Cloud Browser**: Free for a limited time.

## Use cases

BrowserAct is designed for data tasks that require a real browser rather than static page requests alone:

### Product and supplier research

Track products, prices, ratings, promotions, sellers, suppliers, and minimum order quantities across catalogs and marketplaces.

### Customer and competitor research

Collect posts, comments, reviews, pricing, product updates, and source URLs for market and customer analysis.

### Lead generation and business data

Build business lists from browser-accessible directories and websites, including names, categories, locations, ratings, websites, and publicly available contact details.

### Creator, product launch, advertising, and job market data

Track creators, product launches, advertisements, and job listings across public platforms.

### Authenticated dashboards and account workflows

Reuse authenticated browser sessions to open dashboards, operate controls, complete multi-step workflows, and export reports or files.

### Batch and multi-account operations

Run sessions in parallel with rotating identities for batch collection, or maintain stable browser profiles and regional routes for account-based workflows.

Target websites must be accessible through a browser. Users are responsible for complying with target website terms, applicable laws, and access permission requirements.

## Getting started with BrowserAct

### Start a scraper in the cloud

1. Open the [BrowserAct homepage](https://www.browseract.com/).
2. Describe the data you need, the target website, and the expected output.
3. Let the Agent build and test the scraper Bot on the live website.
4. Run the generated Bot and receive structured results and run logs.
5. Refine the prompt when your requirements evolve.

### Give a local AI Agent browser capabilities

1. Open [AI Agents CLI](https://www.browseract.com/cli).
2. Read the [Agent CLI documentation](https://docs.browseract.com/agent-cli/).
3. Install the BrowserAct skill in the Agent environment. The command currently shown on the website is:

   `npx skills add browser-act/skills --skill browser-act`

4. Ask the Agent to navigate websites, interact with pages, and extract the required data.

### Connect existing systems

1. Review the [API reference](https://docs.browseract.com/workflow/api-reference).
2. For MCP clients, use the [MCP integration guide](https://docs.browseract.com/integrations/mcp).
3. Connect published Bots to [Make](https://docs.browseract.com/integrations/make), [n8n](https://docs.browseract.com/integrations/n8n), Zapier, or other application workflows.

## Frequently asked questions

### What is BrowserAct best at?

BrowserAct is best suited to repeatable web data tasks that require a real browser, including dynamic pages, navigation, authenticated sessions, interactive controls, protected pages, and structured output. It can turn natural-language requirements into reusable Bots or provide browser capabilities directly to local AI Agents.

### Do I need to write scraper code or maintain selectors?

No, not for prompt-driven cloud workflows. Describe the data you need in natural language, and BrowserAct builds and tests a reusable scraper. Developers who need greater control can use the CLI, API, workflows, or MCP.

### Should I use Cloud Runtime or AI Agents CLI?

Use Cloud Runtime when you need managed execution from anywhere without local installation or computer storage. Use AI Agents CLI when a local coding Agent needs direct browser control, access to authenticated local Chrome sessions, or interactive operation.

### Can BrowserAct handle login and CAPTCHA workflows?

BrowserAct supports authenticated browser sessions, common CAPTCHA handling, and remote human assistance for steps such as 2FA that require user action. Actual support depends on the website flow and the user's access permissions.

### What happens when a website changes?

When a page changes during a run, the Agent can adapt to the current state. If business requirements change or the website structure changes significantly, refine the scraper Bot through prompts, retest it, and publish the updated workflow rather than assuming the previous path will remain valid.

### What can a single run output?

A run can output structured web data, files, downloads, and run logs. The exact result depends on the task and the output requested in the prompt.

### How are BrowserAct plans and usage billed?

Subscription plans include monthly Credits and usage limits. Browser operations and proxy services consume Credits according to the current pricing page. Visit the [BrowserAct pricing page](https://www.browseract.com/pricing) for current plan amounts, promotions, and allowances.

### Is web scraping legal?

Users should review target website terms, access permissions, and applicable laws before collecting data. BrowserAct provides browser automation infrastructure and does not determine whether a specific scraping task is legally permitted.

## Contact and support

- **Product support**: [support@browseract.com](mailto:support@browseract.com)
- **Documentation**: [Agent CLI documentation](https://docs.browseract.com/agent-cli/), [Workflow API reference](https://docs.browseract.com/workflow/api-reference), and [MCP integration guide](https://docs.browseract.com/integrations/mcp)
- **Community**: [Discord](https://discord.gg/UpnCKd7GaU)
- **Source code and skills**: [GitHub](https://github.com/browser-act/skills)
- **Updates**: [Blog](https://www.browseract.com/blog), [X](https://x.com/browseract), [LinkedIn](https://www.linkedin.com/company/browseract/posts/?feedView=all), and [YouTube](https://www.youtube.com/@browseract)
- **Legal**: [Terms of Service](https://www.browseract.com/terms-of-service) and [Privacy Policy](https://www.browseract.com/privacy-policy)

Official website: https://www.browseract.com/