docker/compose-for-agents

What it solves

This project provides a collection of ready-to-use templates for deploying AI agent systems. It removes the complexity of configuring environments by using Docker Compose to manage the dependencies, models, and tools required for various agent frameworks.

How it works

Each demo is a self-contained folder that includes a compose.yaml file. You simply navigate to the folder, provide any necessary API tokens in a configuration file, and run docker compose up --build. The setup can run models locally using Docker Model Runner (leveraging your GPU) or be configured to use OpenAI models instead.

Who it’s for

Developers who want to experiment with different AI agent frameworks—such as CrewAI, LangGraph, or Agno—without spending time on manual infrastructure setup. It is ideal for those who want to see how agents interact with tools (MCPs) and local or remote language models in a standardized environment.

Highlights

• Includes a wide variety of agent frameworks like CrewAI, LangGraph, Agno, and ADK.
• Supports both local model execution (via GPU) and cloud-based models like OpenAI.
• Standardizes the deployment process using familiar Docker Compose commands.
• Demonstrates integration with various tools and data sources via Model Context Protocol (MCP).
• Provides a consistent way to test multi-agent and single-agent architectures.