Has anyone used docker with ADK with MCP servers?

[xuanyang15]

Hi 👋

I have gotten MCP servers working with ADK but I am trying to figure out to make it work using docker so that the ADK app can be deployed anywhere.

From what i understand now, best practice would be to use docker compose so that each MCP server is a separate container.
I also found out that docker desktop has their own MCP Toolkit now which seems to be great.

Search online, this depo, and the subreddit, docker does not really seem to be something that is used right now when it comes to ADK. Or it could just be that people that are learning it (like me) does not have the skills yet to implement it.

[xuanyang15]

Response from ADK Answering Agent

TLDR: Yes, using Docker with ADK is a recommended practice for creating portable and scalable applications. While the documentation doesn't provide a specific Docker Compose example, the recommended architecture for using ADK with MCP servers lends itself perfectly to a multi-container setup using Docker Compose. You can find examples of how to containerize an ADK agent using a Dockerfile in the Cloud Run and GKE deployment guides.

---

Hi there! That's an excellent question. Using Docker to containerize your ADK application, especially when dealing with MCP servers, is a great approach for ensuring consistency and simplifying deployments. Here’s a breakdown based on the documentation:

Containerizing Your ADK Agent

The official ADK documentation confirms that ADK is designed to be "Deployment Ready" and can be containerized with Docker for deployment on platforms like Cloud Run, GKE, or any other environment that supports containers. [3, 4, 7]

The deployment guides for Google Cloud Run [3] and GKE [7] are your best reference here. They provide concrete examples of the necessary files:

1. Dockerfile: A standard Python Dockerfile that copies your agent code, installs dependencies from requirements.txt, and sets up the environment. The GKE example shows a good starting point. [7]
2. requirements.txt: This file should include google-adk and any other Python dependencies your agent needs. [7]
3. main.py: An entry-point script that initializes and runs your agent, typically using the FastAPI server that comes bundled with ADK. [7]

Integrating MCP Servers with Docker Compose

You are correct that Docker Compose is the ideal tool for managing a multi-service setup. As the documentation on MCP tools shows, an ADK agent often acts as a client to one or more MCP servers running as separate processes. [5, 6]

While the documentation doesn't provide a ready-made docker-compose.yml file, you can structure one as follows:

• A service for your ADK Agent: This service would be built using the Dockerfile mentioned above.
• A service for each MCP Server: Each MCP server your agent communicates with would be its own service in the Compose file. For instance, if you're using the MCP File System Server, you might run it in a container using the npx command shown in the docs. [6]

Here is a conceptual docker-compose.yml example to illustrate the idea:

version: '3.8'
services:
  adk_agent:
    build:
      context: . # Assumes Dockerfile is in the root
      dockerfile: Dockerfile
    ports:
      - "8080:8080"
    environment:
      - GOOGLE_API_KEY=${GOOGLE_API_KEY}
      # Other necessary environment variables

  mcp_filesystem_server:
    image: node:20-slim # Or a more specific image if available
    command: >
      sh -c "npx -y @modelcontextprotocol/server-filesystem --root /data"
    volumes:
      - ./my-files:/data # Mount local files into the container

In this setup, your ADK agent, running in the adk_agent container, would be configured to connect to the mcp_filesystem_server service. This creates a self-contained, portable development and deployment environment.

[1] https://github.com/google/adk-docs/blob/main/docs/streaming/custom-streaming.md
[3] https://github.com/google/adk-docs/blob/main/docs/deploy/cloud-run.md
[4] https://github.com/google/adk-docs/blob/main/docs/index.md
[5] https://github.com/google/adk-docs/blob/main/docs/mcp/index.md
[6] https://github.com/google/adk-docs/blob/main/docs/tools/mcp-tools.md
[7] https://github.com/google/adk-docs/blob/main/docs/deploy/gke.md