Harbor: One MCP Gateway for Many Services with Runtime Skills
Link to open source: https://github.com/vijaydeepsinha/harbor
Harbor is an open-source MCP (Model Context Protocol) gateway that helps AI clients like Claude, Cursor, and ChatGPT work with existing backend APIs without requiring changes to those APIs.
The project started from a simple problem: exposing every API endpoint as an MCP tool doesn't scale very well in real-world systems. As the number of services grows, tool counts grow too, making onboarding and discovery harder for both developers and AI agents.
Harbor takes a different approach. Instead of creating a tool for every endpoint, it provides a small set of reusable tools, runtime skills, and a code execution layer that allows agents to discover services, understand business context, and interact with APIs more effectively.
Some of the key features include runtime skills, capability-driven onboarding, secure sandboxed execution, OAuth 2.1 support, audit logging, and production-focused operational controls.
The goal is simple: make it easier to connect AI systems to existing enterprise services without creating hundreds of tools or rewriting existing APIs.
