Awesome MCP › Browser Automation

Skyvern-AI/skyvern

⭐ 22994 Python added to this list on 2025-05-06 repository created 2024-02-28

Skyvern is a cutting-edge platform that automates browser-based workflows by leveraging large language models (LLMs) and computer vision technologies. It offers a robust, adaptable, and scalable solution for web automation that does not rely on brittle selectors like XPath or DOM parsing, which are prone to breaking when website layouts change. Instead, Skyvern uses Vision LLMs to understand and interact with websites visually, enabling it to operate effectively even on websites it has never encountered before. This approach allows Skyvern to map visual elements to necessary actions to complete workflows without requiring customized code for each site. Skyvern's architecture is inspired by task-driven autonomous agent designs such as BabyAGI and AutoGPT, but it enhances these concepts by integrating browser automation libraries like Playwright. It employs a swarm of agents to comprehend websites, plan actions, and execute tasks, making it highly resilient to website layout changes and capable of applying a single workflow across multiple websites. This makes it particularly useful for complex web automation tasks such as filling out forms, logging in, downloading files, and competitor analysis. The platform provides a simple API and a user-friendly UI for running tasks, either locally or via the managed Skyvern Cloud service, which includes features like anti-bot detection, proxy networks, and CAPTCHA solvers. Skyvern supports advanced usage scenarios, including controlling local or remote browsers and ensuring consistent output schemas. Skyvern has demonstrated state-of-the-art performance on benchmarks like WebBench, especially excelling in WRITE tasks relevant to robotic process automation (RPA). Its ability to reason through interactions using LLMs allows it to handle complex scenarios, such as inferring answers based on context or recognizing product equivalences despite minor differences. Overall, Skyvern represents a significant advancement in web automation technology by combining AI-driven vision and language understanding with browser automation, making it a powerful tool for automating diverse and complex web workflows without the fragility of traditional methods. It also integrates with the Model Context Protocol (MCP), aligning it with modern AI ecosystem standards.

https://github.com/Skyvern-AI/skyvern

ai-driven-automationai-driven-web-interactionanti-bot-detectionapiautomationautonomous-agentsbrowserbrowser-automationcaptcha-solverscomputercomputer-visiondynamic-auto-completegptlarge-language-modelsllmmcpmodel-context-protocolmulti-agent-systempassword-managementplaywrightproxy-networkpythonrobotic-process-automationrpaskyvern-cloudtask-driven-agentstwo-factor-authenticationvisionvision-llmsweb-automationweb-scrapingwebbench-benchmarkworkflowworkflow-automation