A new MCP server slashes AI browser automation costs by 99%

A developer has released turbo-browser-mcp, a new Model Context Protocol server built on Playwright designed to dramatically reduce the cost and latency of automating websites with large language models. The system consolidates dozens of browser primitives into just ten batched tools that can handle complex tasks like pre-filling forms, uploading files from disk, and managing multiple tabs in a single turn instead of requiring multiple back-and-forth interactions with an AI.

The performance gains are substantial: the server achieves 93-99% reductions in token usage compared to standard methods, cutting the cost of AI operations. For example, uploading a 48KB PDF file uses just 85 tokens instead of nearly 45,000 by avoiding base64 encoding. A 10-field complex form that normally requires 11 turns with an AI can now be completed in just one turn, making browser automation with LLMs faster and far more economical for developers building AI agents that interact with web interfaces.