Runtime Deployment

Overview

Deploy Docker images to your runtime environment with Fly:

  • Find the right release using semantic search
  • Create image pull credentials for Fly Registry
  • Choose your method: Agentic (IDE) or Manual (tokens)

Runtime Deployment Flow

To deploy with Fly Registry:

  1. Find the right release using semantic search (Fly Chat or IDE):

    • “Find the release where I fixed the login bug”
    • “Show me the last successful release”
    • “Which release added user authentication?”

  2. Create an image pull secret for Fly Registry authentication

  3. Deploy using the image path format: <your-fly-subdomain>.jfrog.io/docker/<image-name>:<tag>

Example: Deploy to Kubernetes with Fly

Step 1: Generate an access token

Create a token in the Fly Web App for Kubernetes to authenticate with Fly Registry. When you create the token, you’ll receive both a token and a token-specific username.

Create an access token →

Step 2: Configure Kubernetes

Create an image pull secret with your Fly Registry credentials:

kubectl create secret docker-registry <secret-name> \
  --docker-server=<your-fly-subdomain>.jfrog.io/docker \
  --docker-username=<your-fly-username> \
  --docker-password=<your-fly-token> \
  --namespace=<namespace>

Note: Replace <secret-name> with any name you choose (e.g., fly-registry-secret). Use <your-fly-username> and <your-fly-token> from Step 1.

Use the Fly Registry image path in your deployment: <your-fly-subdomain>.jfrog.io/docker/<image-name>:<tag>

Example: acmecorp.jfrog.io/docker/payment-service:v2.3.1

Two Deployment Methods

Method 1: Agentic from your IDE

Use natural language commands in your IDE with Fly MCP assistance.

Prerequisites:

  • Fly Desktop App is installed and running
  • Agentic IDE (Cursor or VS Code Copilot) is connected
  • kubectl configured for your cluster

Example Prompts:

  • “Deploy release v2.3.1 to production”
  • “Deploy payment-service to staging namespace”
  • “Find the release with the login fix and deploy to staging”

Fly MCP automatically handles image pull secret creation and deployment configuration.

Method 2: Manual

Use access tokens and standard deployment tools like kubectl or Helm. Ideal for CI/CD pipelines, automated scripts, or environments without IDE access.

Steps:

  1. Create a token in Fly Web with read-only permission (learn more →)

  2. Create the image pull secret using the kubectl command shown above with your token

  3. Deploy using your tool with image path: <your-fly-subdomain>.jfrog.io/docker/<image-name>:<tag>

Next Steps

Now that you understand the developer workflow, explore the tools to manage your team and artifacts: