Sentinel screenshot

What is Sentinel?

Sentinel is an open-source AI browser automation tool built on Playwright that uses large language models to control web browsers. It reduces token consumption by using more efficient models like Gemini Flash instead of costlier alternatives, making it significantly cheaper to run than comparable tools. The tool is designed for developers and teams who need to automate browser interactions, web testing, or data extraction tasks without incurring high API costs. It includes features like self-healing locators that adapt when page layouts change, parallel execution for running multiple automation tasks simultaneously, and OpenTelemetry integration for monitoring and observability.

Key Features

LLM-powered browser automation

Uses AI models to understand and interact with web pages naturally, rather than relying on brittle selectors

Self-healing locators

Automatically adapts to minor changes in page structure without requiring manual updates

Parallel execution

Run multiple automation tasks concurrently for faster batch processing

Token-efficient

Uses Gemini Flash and similar models to reduce API costs compared to GPT-4o alternatives

OpenTelemetry support

Built-in observability and monitoring for tracking automation performance and debugging

Open-source

Code is publicly available, allowing inspection, modification, and self-hosting

Pros & Cons

Advantages

  • Significantly lower operating costs due to token-efficient model selection
  • Self-healing capabilities reduce maintenance burden when websites update
  • Parallel execution improves throughput for large-scale automation tasks
  • Open-source nature provides transparency and flexibility for customisation

Limitations

  • As an open-source project, support may be community-driven rather than commercial
  • Reliance on LLM performance means results can vary based on model quality and page complexity

Use Cases

Web scraping and data extraction from dynamic websites without writing brittle CSS selectors

Automated testing of web applications using natural language instructions

Monitoring and alerts for changes across multiple websites

Form filling and data entry automation at scale

Screenshot and content collection for competitive analysis or archival