browserable: Automate Browsers for AI Agents

Overview

Browserable lets AI agents interact with websites by navigating pages, clicking controls, filling forms, and extracting information. It is designed for developers building browser-based agents who want a browser automation service they can run themselves or integrate through an API.

The project combines an agent task interface with configurable LLM and remote-browser providers. It may suit teams that need to automate multi-step web tasks, while still choosing their own provider configuration.

Key Features

• Runs browser tasks described in natural language.
• Supports navigation, form entry, button clicks, and information extraction.
• Includes a JavaScript SDK for creating tasks and waiting for results.
• Provides a REST API for integrating task execution into other applications.
• Can be self-hosted, with a setup flow available through npx browserable.
• Offers configuration for LLM providers, storage, databases, remote browsers, and custom functions.

Use Cases

• Researchers can collect details from websites or locate recent papers without manually repeating searches.
• Developers can add browser-based research or data collection tasks to AI agent applications using the SDK or REST API.
• Teams can automate multi-step shopping or comparison searches where results depend on several website criteria.
• Operators who need control over deployment can self-host the service and configure their own LLM and remote browser providers.

Getting Started

Run the interactive setup:

npx browserable

Then visit http://localhost:2001 to configure LLM and remote browser API keys. See the README and documentation for setup details and SDK and REST API guides.

Alternatives

• pinchtab: PinchTab exposes direct Chrome control through a local Go server and CLI, while Browserable runs natural-language tasks through a self-hostable library and REST API.
• magentic-ui: MagenticLite focuses on supervised browser and local-file workflows with human approval, while Browserable provides a self-hostable browser automation library and REST API.
• OpenWebAgent: OpenWebAgent combines a Chrome extension with a configurable server for developers supplying their own models, while Browserable offers a JavaScript SDK and REST API for natural-language tasks.
• brightdata-mcp: Bright Data MCP provides managed web data and remote browser tools via an API token, while Browserable is a self-hostable browser automation library.

Considerations

Browserable requires configured LLM and remote browser providers to run tasks. The documented manual development setup requires Docker and Docker Compose, and self-hosting involves several services, including MongoDB and Redis. The repository describes a provider-based configuration, so users should check the documentation for supported options and deployment requirements.