As LLMs become autonomous actors, MCP defines how they interact with tools, while containers define how they run safely.

In this workshop, using observability tools such as Grafana as an example, you'll learn how to run observability MCP servers as containers via Docker MCP Gateway and use them with sandboxed AI agents to manage dashboards and alerts using natural language, while preserving strong isolation and clear trust boundaries. We'll also show how this local setup maps cleanly to CI/CD pipelines, demonstrating how developers can start locally and scale the MCP-based architecture to Kubernetes.

By the end of the workshop, you'll have:
* A secure local MCP setup running in containers.
* AI agents in sandboxed environments interacting with real tools via Docker MCP Gateway.
* A path how to translate the local MCP experience to Kubernetes using Compose-For-Agents