MCP (Model Context Protocol) connectors bridge the gap between LLMs/agents and tools by standardizing how tools are discovered, invoked, and managed. While MCP provides a standard for exposing tools, connectors handle the complex orchestration of connecting runtimes to MCP servers without requiring custom client logic. Tool-specific implementations create fragmentation issues with inconsistent formats, scattered authentication, and lack of central governance. AI gateways like Portkey solve this by providing unified MCP connectors that route requests across multiple servers, apply consistent authentication, and enforce governance policies across all tool calls.
Table of contents
What MCP connectors solveThe challenges of working with tool-specific MCP connectorsSort: