Enables parallel, isolated web browsing sessions for multiple AI agents through a modified Playwright MCP server.
Sponsored
The official Playwright MCP server uses a single shared browser context, which restricts multiple AI agents from simultaneously browsing the web without state conflicts. This unofficial fork provides a pragmatic solution by introducing browser instance isolation, ensuring each agent operates within its own independent session. It's designed to facilitate multi-agent workflows where parallel and distinct web interactions are crucial, effectively circumventing the limitations of the upstream server until native support for parallel instances is available.
Key Features
01Adds an `instanceId` parameter to all standard Playwright tools for targeting specific browser sessions
02Backward compatibility with standard Playwright MCP when `instanceId` is omitted
03Dedicated tools to create, list, and close browser instances
04Supports both HTTP/SSE and Stdio server modes for flexible integration
05Isolated browser instances for each agent's session
061 GitHub stars
Use Cases
01Running multiple AI agents that require independent web browsing sessions simultaneously
02Developing multi-agent systems where agents perform parallel online tasks without interfering with each other's state
03Testing web applications with multiple concurrent, isolated user flows from a single MCP server
