Docker plans Model Context Protocol security boost

Docker Desktop will soon support a curated catalog of Model Context Protocol servers in Docker Hub and, with a new toolkit, integrate with enterprise developer workflows.

Model Context Protocol (MCP), created by frontier AI model maker Anthropic, has garnered support from rivals such as OpenAI in the last month, along with large IT vendors including Microsoft and Google. Docker Inc. is the latest vendor to pledge support for the protocol, which proposes a standard specification for connecting AI agents to data sources and tools. AI agents are software components backed by large language models that can independently perform tasks and execute workflows.

The Docker MCP Catalog and Docker MCP Toolkit are "upcoming," but no expected availability date was provided, according to a Docker press release issued this week. A Docker spokesperson said they are expected to become available in the next month.

Other vendors, such as Cloudflare, Stytch and Okta subsidiary Auth0, have released identity and access management support for MCP, which doesn't support enterprise-grade access controls on its own. Docker MCP Toolkit will build in registry and image access management controls for the Docker MCP Catalog, a set of curated MCP servers built on Docker Hub, along with pluggable support for secrets management tools such as HashiCorp Vault.

What is different about Docker is that the Docker MCP server executes isolated code in Docker containers, supporting multi-language scripts, dependency management, error handling and container lifecycle operations.

"Almost everyone is trying to deploy MCP servers and an MCP catalog as quickly as possible," said Andy Thurai, independent analyst at The Field CTO. "What is different about Docker is that the Docker MCP server executes isolated code in Docker containers, supporting multi-language scripts, dependency management, error handling and container lifecycle operations."

This could be useful to developers who need secure, isolated environments for executing untrusted or experimental code, Thurai said. The news comes as security researchers uncover potentially serious vulnerabilities in the protocol without third-party hardening support. Researchers from AWS and Intuit have proposed a zero-trust security framework for it.

And so far, MCP is at an experimental stage. It remains under Anthropic's governance, although the company has said it will explore donating the project to an open source foundation. It's also quite early in the development of agentic AI. While individual AI agents have become available to perform specific tasks, the underlying infrastructure for agentic AI is still being built, including other protocols that connect agents to one another across orchestration frameworks.

Still, Torsten Volk, an analyst at Enterprise Strategy Group, now part of Omdia, said Docker should establish support for MCP as soon as possible.

Torsten Volk

"While it may be early to jump on the MCP train, it is crucial for Docker to be the first to build out an ecosystem of MCP servers that allows developers to easily add all kinds of tools and data APIs to their applications, without having to worry about security or writing code," he said. "Similar to using Docker Hub as an image registry, using an MCP catalog to add more advanced AI-driven capabilities to applications makes Docker Desktop [stickier]."

This ultimately benefits Docker Desktop users if Docker can draw in third-party MCP servers and make them available using Docker Hub, where developers can easily find and combine them, Volk said.

So far, the Docker MCP Catalog has more than 100 client listings for AI tools such as Docker AI Agent, Anthropic's Claude, and agentic AI integrated development environments such as Cursor, Visual Studio Code and Windsurf. Launch partners include Elastic, Grafana Labs and New Relic.

This list of partners must be fleshed out further for Docker MCP tools to succeed, Thurai said.

"Docker lifecycle management for MCP helps avoid resource leaks and optimize infrastructure costs in production environments, and its multilingual support means it works in any environment with any tool of choice," he said. "However, their partner ecosystem is still very weak. Hopefully, they can bring enough mass to make it compelling to the developer audience they have."

Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.