Deployment¶

FastMCP Runner can be deployed anywhere that runs containers. This section covers platform-specific configuration for common deployment targets.

Choosing an image tag¶

For production deployments, use either latest-stable or a specific version tag:

# Recommended for production
ghcr.io/drengskapur/fastmcp-runner:latest-stable

# For reproducible deployments
ghcr.io/drengskapur/fastmcp-runner:v0.1.0

The latest tag tracks the main branch and may include unreleased changes. Use it only for development and testing.

Registry credentials¶

Most deployments require credentials for pulling your private MCP application image. How you provide these depends on your platform:

• Environment variables: Set REGISTRY_USER and REGISTRY_PASSWORD directly
• Secrets management: Reference secrets from your platform's secret store
• Service accounts: Some platforms support workload identity or service account authentication

See the platform-specific guides for details on secrets configuration.

Health checks¶

FastMCP Runner includes a built-in health check that polls the configured health endpoint:

HEALTHCHECK --interval=30s --timeout=10s --start-period=30s --retries=3 \
  CMD wget -q -O /dev/null "http://localhost:${PORT}${HEALTHCHECK_PATH:-/health}" || exit 1

Configure your orchestrator's health checks to use this endpoint. The default path is /health; change it with the HEALTHCHECK_PATH environment variable if your application uses a different path.

Resource requirements¶

FastMCP Runner itself is lightweight—the base image is under 50MB and the entrypoint adds minimal overhead. Resource requirements depend primarily on your MCP application.

As a starting point:

The startup phase temporarily requires additional disk space for image extraction. Plan for roughly 2x your application image size during startup.

Platform guides¶

• Hugging Face Spaces — Docker Spaces deployment
• Docker Compose — Local and self-hosted deployments