7 Essential Steps to Master Custom MCP Catalogs and Profiles for Enterprise AI

Managing AI tooling at scale just got a whole lot easier. Today, we're diving into two game-changing capabilities—Custom MCP Catalogs and MCP Profiles—that together transform how organizations package, distribute, and use Model Context Protocol (MCP) servers. Whether you're a platform engineer looking to curate trusted tools or a developer wanting to share configurations across teams, this listicle walks you through everything you need to know. From building your own MCP server to creating portable profiles, we cover the practical steps to advance your enterprise MCP adoption. Let's get started.

1. Understanding Custom MCP Catalogs

Custom MCP Catalogs allow organizations to curate and distribute approved collections of MCP servers. Instead of each team member hunting for servers across the open internet, you can publish a centralized catalog that defines which servers are trusted. This includes internally built servers, community sources, and servers from Docker’s own MCP Catalog. The result is a single, governed experience where developers can discover and use vetted AI tools without breaking security or compliance policies. By bringing control, flexibility, and trust together, Custom Catalogs become the foundation for enterprise-grade MCP management.

7 Essential Steps to Master Custom MCP Catalogs and Profiles for Enterprise AI — Source: www.docker.com

2. Harnessing the Power of MCP Profiles

MCP Profiles are a new primitive that let you define portable, named groupings of MCP servers. Think of them as reusable configurations that individual developers can build, run, and share across projects and teams. Profiles solve real-world use cases today: you can bundle a set of servers for a specific workflow, share that profile with teammates via Git, and instantly replicate the same tooling environment everywhere. This reduces setup friction and ensures consistency. Going forward, Profiles will also form the foundation for advanced features like role-based access and automated deployment.

3. Step-by-Step: Building Your Own MCP Server

To create a custom catalog, you first need a custom MCP server. For illustration, we built a reference server called roll-dice (find it on GitHub). It's a standard MCP server that communicates over stdio and is containerized as a Docker image. After building and pushing the image to Docker Hub, you create a metadata YAML file describing the server—its name, title, type, image location, and purpose. For example: name: roll-dice; image: yourdockerhub/mcp-dice@latest; description: An MCP server that can roll dice. This metadata becomes part of your catalog entry.

4. Creating a Custom Catalog That Mixes Sources

Now you can assemble a custom catalog that includes both public servers from the Docker MCP Catalog and your own internally built servers. Using the CLI, you define a catalog JSON or YAML file that references each server's metadata. For instance, you might list the official “filesystem” server alongside your “roll-dice” server. The catalog can be versioned, signed, and stored in a Git repository or Docker Hub. This selective curation ensures teams only access what’s approved. To import the catalog, developers simply run a command like docker mcp catalog import pointing to your catalog file.

Source: www.docker.com

5. Importing Your Catalog via Docker Desktop

For a more visual experience, Docker Desktop supports importing custom catalogs. After your catalog is published (e.g., on Docker Hub or a private registry), users can open Docker Desktop, navigate to the MCP section, and add the catalog URL. The interface then displays all approved servers in a single, searchable list. This lowers the barrier for less technical team members and provides a unified dashboard for managing AI tooling. It’s the perfect way to roll out enterprise-wide MCP adoption without requiring everyone to master the CLI.

6. Combining Profiles with Catalogs for Team Efficiency

The real magic happens when you combine Custom Catalogs with MCP Profiles. A team lead can create a catalog of approved servers, then define a profile that selects a specific subset for a particular project. That profile is stored as a portable file (e.g., mcp-profile.yaml) and can be shared via version control. When a developer clones the repo, they run a single command to apply the profile, instantly activating the correct MCP servers. This eliminates manual setup, reduces errors, and enforces governance without hindering developer velocity.

7. Future Horizons: Scaling Enterprise MCP Adoption

Custom Catalogs and Profiles are just the beginning. Future enhancements will include role-based access control, automated catalog updates, and integration with CI/CD pipelines. Imagine a catalog that automatically updates when a new version of an internal server is built, or a profile that enforces different server sets for QA, staging, and production. As the MCP ecosystem grows, these primitives will become essential for any organization serious about AI tooling governance. By adopting them early, teams lay the groundwork for scalable, secure, and efficient AI development.

Conclusion: Custom MCP Catalogs and Profiles are not just features—they’re the building blocks for a disciplined, enterprise-ready approach to AI tooling. Starting with building your own server and catalog, then layering in portable profiles, you can transform how your team discovers, shares, and manages MCP servers. The steps outlined here give you a clear path to move from chaos to control. Dive in, experiment with the CLI and Docker Desktop, and watch your team’s productivity soar.

Tags: