Show HN: One MCP setup for 22 clients, with lazy tool discovery (github.com)

A new local gateway for AI clients, Toolport, has been introduced to streamline the connection and management of multiple MCP (Model Context Protocol) servers. Instead of configuring each server separately for every AI tool like Claude and Codex, users can set up Toolport once. This drastically reduces the overhead of tool definitions, cutting token costs by approximately 96%—for example, using 900 tokens instead of 24,000 with three interconnected servers. Toolport implements a "lazy discovery" mode where the agent only queries the required tools on-demand, keeping the context flat regardless of the number of connected servers. This development is significant for the AI/ML community as it enhances operational efficiency and reduces costs associated with token usage. Toolport integrates security features, including tool integrity checks to detect potential "rug pulls" or tool poisoning and governance options for controlling tool visibility and usage per client. By simplifying server management and minimizing the computational burden on AI clients, Toolport facilitates scalability in AI deployment while maintaining robust security protocols, making it a valuable tool for developers and organizations employing AI tools.