Skyvern

Skyvern is an MCP server that provides AI applications with programmatic browser control and automated web interaction capabilities. It connects Large Language Models to web environments through either a local server or Skyvern Cloud, enabling agents to navigate pages, inspect visual layouts, and handle dynamic multi-step workflows. Designed for developers and automated workflows, it allows AI clients like Claude Desktop, Cursor, Windsurf, and OpenCode to operate browsers directly rather than relying on brittle scrapers. Through this interface, AI agents can handle tasks such as filling out complex multi-page online forms, downloading remote documents, extracting content from dynamic single-page applications, and gathering web search results. Users can run Skyvern locally using custom language models or delegate browser execution to Skyvern Cloud using an API key. Operating through stdio transport, it exposes dedicated browser session tools to preserve state during web navigation tasks.

Category: Browser & Web Automation

Tags: automation, browser, headless browser, playwright, web scraping

Visit Skyvern

How to install and configure Skyvern

1. Ensure your system is running Python 3.11. 2. Install Skyvern using pip: bash pip install skyvern 3. Run the configuration wizard: bash skyvern init 4. If running locally rather than using Skyvern Cloud, start the local server in a separate terminal: bash skyvern run server 5. To configure an MCP client manually, add the following entry to your MCP client configuration (such as Claude Desktop or Cursor): json { "mcpServers": { "Skyvern": { "env": { "SKYVERN_BASE_URL": "https://api.skyvern.com", "SKYVERN_API_KEY": "YOUR_SKYVERN_API_KEY" }, "command": "PATH_TO_PYTHON", "args": [ "-m", "skyvern", "run", "mcp" ] } } }

What you can do with Skyvern

• Navigating to news aggregators like Hacker News to parse and extract top trending discussions for automated summaries. * Querying job boards directly through an IDE like Cursor to gather local programming positions and compensation data. * Automating government or corporate portal searches, such as Form 5500 lookups, and downloading target filings directly. * Filling out multi-step web forms and interacting with complex interactive user interfaces without hardcoded scraping scripts. * Executing multi-page research workflows where an AI agent inspects dynamic single-page applications to retrieve structured datasets.

Key facts

• https://github.com/Skyvern-AI/skyvern
• Browser & Web Automation, Developer Tools & Code Intelligence
• automation, browser, headless browser, playwright, web scraping

Part of MCP Servers

What is Skyvern?

Skyvern is an MCP server and automation platform that connects AI assistants like Claude to a headless browser. It enables models to interact with websites, fill out forms, download files, and extract web data.

How do I install Skyvern?

You install Skyvern by running pip install skyvern in a Python 3.11 environment. Afterward, run skyvern init to complete the configuration wizard for your selected MCP application or cloud credentials.

Which MCP clients work with Skyvern?

Skyvern explicitly supports Claude Desktop, Cursor, Windsurf, and OpenCode through its setup command, and it can also be configured in any client that supports standard Model Context Protocol servers over stdio.

What Python version does Skyvern require?

Skyvern strictly requires a Python 3.11 environment to run properly. Attempting to install or run the server on other Python versions may lead to dependency or runtime errors.

What is the difference between local mode and Skyvern Cloud?

Local mode runs the Skyvern server and your chosen LLM on your own infrastructure, requiring you to execute skyvern run server. Skyvern Cloud offloads execution to remote browser sessions managed via an API key.