Web-MCP-Bridge – Connect browser-based LLMs to MCP servers and tools (github.com)

The newly announced Web-MCP-Bridge enables seamless integration between browser-based large language models (LLMs) like ChatGPT, Gemini, and Claude, and local Model Control Protocol (MCP) servers, eliminating the need for API keys or extra subscriptions. This tool allows users to enhance free AI chat interfaces by adding the capability to run commands on their local MCP servers via WebSocket. Whenever a model requires a tool, it sends a JSON request that is executed locally, with users granting approval for each operation to ensure security. This development is significant for the AI/ML community as it enhances the interactivity and functionality of LLMs by allowing them to utilize local resources directly, promoting more personalized and powerful applications without compromising user data security. Technically, the architecture mitigates prompt injection risks by presenting the results from MCP calls as data rather than instructions. The project is open source, inviting community contributions for new adapters and improvements, ultimately fostering collaboration and innovation among developers.