10 Key Insights Into Custom MCP Catalogs and Profiles for Enterprise AI Tooling

Welcome to the era of organized AI tooling. With the general availability of Custom MCP Catalogs and Profiles, teams can finally tame the chaos of scattered MCP servers. This article unpacks the top ten things you need to know about these two complementary capabilities—from why they matter to how you can start using them today. Whether you're an enterprise architect, a developer, or a platform engineer, these insights will help you package, distribute, and manage your AI tools with confidence.

1. The General Availability Launch

We are thrilled to announce the general availability of Custom Catalogs and Profiles for managing Model Context Protocol (MCP) servers. These two features fundamentally change how organizations package, distribute, and manage AI tooling. Custom MCP Catalogs let teams curate and distribute approved collections of MCP servers, while MCP Profiles enable individual developers to easily build, run, and share their MCP tools and configurations across projects and teams. This launch marks a significant step toward enterprise-grade AI infrastructure.

10 Key Insights Into Custom MCP Catalogs and Profiles for Enterprise AI Tooling — Source: www.docker.com

2. Why Enterprises Need Custom Catalogs

As organizations adopt MCP, a consistent need emerges: teams must curate a trusted list of MCP servers, including internally built ones. Without a central catalog, developers waste time searching across the open internet for servers that may be insecure or unmaintained. Custom Catalogs solve this by allowing IT and platform teams to publish approved server lists. This ensures everyone uses only verified, enterprise-safe MCP servers, reducing risk and improving developer productivity.

3. What Custom MCP Catalogs Are

A Custom MCP Catalog is a curated collection of MCP server definitions. It can reference servers from Docker's MCP Catalog, community sources, or custom servers developed internally. Each server entry includes metadata such as name, description, type (server), and the Docker image location. By combining flexibility, control, and trust in one experience, Custom Catalogs enable centralized discovery of approved AI tools across organizational boundaries.

4. The Power of MCP Profiles

MCP Profiles are a new primitive that lets you define portable, named groupings of MCP servers. Think of them as reusable configurations that bundle together multiple servers for a specific use case or environment. Profiles make it easy for developers to switch between different tool sets—for example, a profile for data analytics and another for coding assistants—without manual reconfiguration. They also simplify sharing across teams, as profiles can be versioned and distributed alongside catalogs.

5. Building Your First Custom MCP Server

To populate a catalog, you need MCP servers. The reference roll-dice server (available on GitHub) demonstrates the basics. It is a standard MCP server communicating over stdio, built as a Docker image and pushed to Docker Hub. The server's metadata is saved in a YAML file (e.g., mcp-dice.yaml) containing fields like name, title, type, image, and description. This metadata is what catalogs consume to present servers to users.

6. Creating a Custom Catalog Step by Step

Creating a catalog involves combining servers from multiple sources. For example, you can include the roll-dice server alongside servers from the Docker MCP Catalog. Write a YAML file that lists server entries—either inline or by referencing external catalog URLs. Once defined, you can publish the catalog to a registry like Docker Hub or a private repository. The result is a single, shareable definition that developers can import with one command.

Source: www.docker.com

7. Using Docker Desktop to Import Catalogs

Docker Desktop provides a user-friendly interface for importing Custom Catalogs. After publishing your catalog, users can point Docker Desktop to its location (e.g., a URL or local file). The app automatically fetches and displays the available servers, allowing developers to browse, select, and enable MCP servers directly from the GUI. This lowers the barrier for non-CLI users and integrates seamlessly into existing Docker workflows.

8. CLI vs. Docker Desktop: When to Use Each

Both the CLI and Docker Desktop offer full catalog functionality, but each shines in different scenarios. The CLI is ideal for automation, CI/CD pipelines, and power users who want scriptable control. Docker Desktop, on the other hand, is designed for everyday developers who prefer a visual interface. For enterprise rollouts, the CLI enables centralized management, while Docker Desktop simplifies adoption across diverse teams. Choose the tool that matches your workflow.

9. Practical Use Cases for Profiles

Profiles solve several real-world problems today. A data science team can define a profile with SQL, visualization, and machine learning MCP servers. A frontend team might create a different profile for code generation and API testing. Profiles can be named (e.g., "data-analytics") and shared via catalogs, enabling consistent tooling across projects. They also make it easy to experiment—switch profiles to test new servers without affecting existing configurations.

10. The Future Foundation for MCP Tooling

Custom Catalogs and Profiles are not just a solution for today—they lay the groundwork for future expansions. We envision features like dynamic profile swapping, policy-based access control, and integration with secrets management. By standardizing how MCP servers are defined, discovered, and grouped, these primitives will support advanced scenarios such as automated deployment, compliance auditing, and multi-cloud tool orchestration. The journey has just begun.

Custom MCP Catalogs and Profiles together deliver a robust framework for enterprise AI tooling. They bring order to the MCP ecosystem, enabling teams to discover, share, and manage servers with confidence. Start building your own catalogs and profiles today, and join us in shaping the future of AI infrastructure.

Tags: