MCP Harbor – A Model Context Protocol Server for Harbor Registries

MCP Harbor bridges the gap between AI assistants and Harbor, a widely used container registry. By exposing Harbor’s REST API as MCP tools, the server lets an assistant perform project, repository, tag, and Helm chart operations directly from a conversation. This eliminates the need for developers to write custom integrations or remember Harbor’s CLI syntax, enabling quick automation and data retrieval within a single AI workflow.

The server itself is a lightweight Node.js application written in TypeScript, which guarantees type safety for the tool definitions and eases maintenance. When launched it listens on a configurable port and presents a catalog of tools such as , , or . Each tool is a declarative wrapper around Harbor’s endpoints, translating input parameters into authenticated HTTP requests and returning structured JSON. The result is a consistent interface that any MCP‑compatible client can consume, regardless of language or platform.

Key capabilities include:

• Full Harbor CRUD – Create, read, update, and delete projects, repositories, tags, and Helm charts.
• Parameter validation – Tool signatures enforce required fields, reducing runtime errors in assistant calls.
• Secure authentication – Credentials are supplied via command‑line arguments or environment variables, ensuring that the server never exposes secrets in logs.
• Debug mode – Optional verbose logging helps developers trace request/response cycles during development or troubleshooting.

Typical use cases are:

• Continuous Delivery pipelines: An assistant can automatically prune unused tags or move images between projects based on policy, integrating seamlessly into CI/CD scripts.
• Security scanning: By listing all tags and feeding them to a vulnerability scanner, an assistant can report exposed images without manual intervention.
• Infrastructure as Code: When provisioning new environments, an assistant can create the necessary Harbor projects and Helm charts on demand, keeping registry state in sync with deployment descriptors.

Integrating MCP Harbor into an AI workflow is straightforward. A developer configures the assistant’s MCP client to point at the server’s URL, then invokes any of the exposed tools as part of a prompt. The assistant can chain multiple calls—first , then if a target is missing, and finally push an image—while maintaining context across the conversation. This tight coupling between registry operations and natural language makes Harbor management more accessible to non‑technical stakeholders.

What sets MCP Harbor apart is its pure protocol focus. It does not add a new API layer or require modifications to Harbor itself; it simply translates existing REST calls into MCP tools. This lightweight, drop‑in design allows teams to adopt it with minimal friction while gaining powerful AI‑driven registry automation.