Build a MCP Server to Control Docker with GitHub Copilot Agent

Imagine commanding your Docker containers right from VS Code, with GitHub Copilot’s Agent mode responding to natural language prompts. With a Model Context Protocol (MCP) server in Python, this becomes reality! it’s AI as your infrastructure interface. In this guide, you’ll build a simple yet powerful MCP server to connect Copilot to Docker, complete with MCP standard details and a dash of LinkedIn-ready enthusiasm. Boost productivity, transform your workflow, and stay relevant in the market.

Why opt for an MCP server when GitHub Copilot Agent can run commands such as "docker ps" or even "docker compose up" directly in the terminal? An MCP server delivers: direct Docker API integration for reliable JSON responses, custom commands (e.g., "stop nginx containers" or "list active containers"), scalability for multiple tools, and native Copilot integration, bypassing fragile text parsing for precise programmatic control. It also enables context-aware, multi-step command chaining (e.g., "build the image, then run it"), secure command validation, and a consistent natural language interface. By returning structured data rather than raw terminal output, MCP servers make it easier for AI agents to understand and act, while offering a pluggable architecture ready for expansion beyond the CLI.

Step-by-Step: Building Your MCP Server

Step 1: Set Up Your Environment

Prep your workspace like a chef laying out ingredients.

• Requirements: Python 3.8+ (python --version). Docker running (docker ps). VS Code with GitHub Copilot and chat.agent.enabled set.
• Install Dependencies:

pip install docker        

Ensure Docker Desktop is running on Windows.

Step 2: Understand the MCP Standard

The Model Context Protocol (MCP) is an open standard for integrating AI models like Copilot with external tools, defined by:

• Communication: HTTP POST with JSON for commands and responses.
• Flexibility: Supports dynamic tools (e.g., Docker operations).
• Interoperability: Standardizes data, avoiding custom APIs.
• Extensibility: Enables local/remote servers with live updates.
• Security: Includes origin validation.

Our server handles JSON commands (e.g., "list_containers") via HTTP, executes Docker actions, and returns formatted responses, as shown in the code comments.

Step 3: Write the MCP Server Code

This server bridges Copilot Agent to Docker, with comments highlighting MCP compliance.

Create mcp_docker_server.py:

MCP Highlights in Code:

• HTTP/JSON: do_POST processes POST with JSON payloads.
• Standardized Responses: JSON with result or error and HTTP codes.
• Dynamic Commands: Handles list_containers, get_container_status, etc.
• Fixed Port: Runs on localhost:8000 for tool registration.

Step 4: Run the Server

Fire up the server!

• In PowerShell:

python mcp_docker_server.py        

• Expected Output:

Starting MCP server...
Attempting Docker connection...
Docker connection successful!
🚀 MCP server running at http://localhost:8000        

• Keep the terminal open.

Step 5: Test the Server

Verify it responds like a true MCP server.

• In another PowerShell:

curl -X POST http://localhost:8000 -H "Content-Type: application/json" -d '{\"command\": \"list_containers\"}'        

• Expected:

{"result": ["test_nginx", ...]}        

Or {"result": []} if no containers exist.

• Create a test container:

docker run -d --name test_nginx nginx        

Retry the curl.

Step 6: Connect to VS Code

Link it to Copilot Agent for magic.

• In VS Code:
• Open Chat (Ctrl+Alt+I).
• Click "Tools" icon > Add:

Name: DockerMCP

Type: HTTP

URL: http://localhost:8000

• Select "Agent" in Chat.
• Type: "Use DockerMCP to list active containers."
• Expected: Container list appears in Chat.

Step 7: Expand Your Horizons

Add custom commands, like:

• start_container:

Pro Tips

• Test Containers: Run docker run -d nginx.
• Logs: Leverage print statements for debugging.
• Port Conflicts: Switch to 9000 if 8000 is busy (server_address = ('localhost', 9000)).
• Security: Add HTTPS for production use.

Conclusion: You’re the Docker Maestro!

With this MCP server, you’ve supercharged GitHub Copilot Agent to command Docker with precision. Run python mcp_docker_server.py, watch for the "🚀", and control containers effortlessly. Test it, expand it, and rule your digital domain!

Here a a fully expanded version of the mcp_docker_server.py script.

Read More

Agent mode available to all users and supports MCP: https://shorturl.at/YadC5

#Python #Docker #AI #GitHubCopilot #VSCode #DevOps #Automation #Coding #TechTips #SoftwareDevelopment #OpenSource #Programming #TechInnovation #DeveloperTools #MachineLearning #MCP