Browserable is an open-source, self-hostable library for building browser automation agents powered by AI. It enables agents to navigate websites, interact with elements, and extract data, targeting developers building AI-driven web interaction tools. The library claims 90.4% on Web Voyager benchmarks.

How It Works

Browserable leverages a microservices architecture orchestrated via Docker Compose, comprising UI, task management, database (MongoDB), caching (Redis), and object storage (MinIO). It integrates with various LLM providers (OpenAI, Gemini, Claude) and remote browser services (Hyperbrowser, Steel) to interpret user tasks and execute browser actions.

Quick Start & Requirements

• Install/Run: npx browserable for guided setup.
• Prerequisites: Docker, Docker Compose.
• Configuration: Requires API keys for an LLM provider and a remote browser service.
• Links: Website, Documentation, Discord.

Highlighted Details

• High benchmark performance on Web Voyager.
• Supports multiple LLM providers and remote browser services.
• Offers a JavaScript SDK for programmatic integration.
• Extensible with custom functions and configurable backend services.

Maintenance & Community

• Active development with a Discord community for support and feedback.
• Open to contributions via pull requests.

Licensing & Compatibility

• The repository does not explicitly state a license in the provided README.

Limitations & Caveats

• The license is not specified, which may impact commercial use or integration into closed-source projects.
• Requires significant setup with Docker and external API keys for core functionality.