Last month, we introduced Hyphen Agent. This month, Agent becomes more capable and more interactive: you can now prompt it on demand, and it has a new skill to optimize your deployment environments based on real workload data.

Below is a summary of what's shipped since the March update. Each item links to the full changelog entry for the details.

Hyphen Agent Updates

Prompt Hyphen Agent

Agent skills normally run automatically based on your organization's policies. With this release, you can also trigger them manually. This is useful when you want to clean up resources before a demo, run a stale flag sweep ahead of a release, or scale infrastructure for a known upcoming event.

Three skills are available on demand:

  • Cloud resource cleanup: specify an "archived prior to" date and have Agent remove unused infrastructure from previous deployments.

  • Stale feature flag removal: trigger flag detection whenever you want a sweep, not only on the automated cadence.

  • Schedule Event: prep your deployment for a traffic spike by scaling ahead of a launch, sale, or marketing moment. Events can be one-off or recurring.

Read more

Agent Can Optimize Your Deployment Environment

Right-sizing infrastructure is usually manual and reactive. Prompt Agent to optimize a project environment, and it creates an optimizer task that analyzes real workload behavior (CPU, memory, and requests per second) and produces ranked configuration recommendations. That includes suggested updates to min/max instance counts and traffic region placement, scoped to the provider and region where the environment actually runs.

Each recommendation carries a risk level. Lower-risk changes may apply automatically; higher-risk ones require approval before execution. The result is less cloud spend on unused capacity, better responsiveness during spikes, and infrastructure that adapts to real regional demand patterns. Read more

Platform Improvements

Deployment Previews

Spin up ephemeral, fully isolated environments from any deployment configuration. Each preview gets its own subdomain, provisions its own infrastructure, and is torn down automatically when deleted. Useful for validating feature branches or sharing working demos without touching staging. Read more

Project Environment Types and One-to-One Deployment Policies

Environments now have explicit types (development, production, or custom) and map one-to-one with deployment policies. The result is a clearer separation between lifecycle stages and more predictable deployment behavior across projects. Read more

Project UX Updated Around Environments

The Project experience now centers on environments as the primary configuration surface. Environment selection drives the resource architecture view and the deployment settings you see, making it easier to navigate between development, production, and custom setups. Read more

Tips

Using Hyphen in GitHub Actions

Hyphen ships three GitHub Actions you can use individually or compose into a single deploy pipeline:

  • Hyphen/setup-hx-action installs and authenticates the hx CLI on the runner so subsequent steps can use it.

  • Hyphen/env-action pulls Hyphen ENV secrets for the chosen environment and writes them to .env files or exports them as environment variables.

  • Hyphen/deploy-action runs a Hyphen deployment, waits for the run to finish, and exposes the deployment ID, run ID, URL, and status as step outputs.

Chained together, you can pull secrets, deploy, and report the result back to the workflow log in just a few steps:

- uses: actions/checkout@v4
- uses: Hyphen/setup-hx-action@v1
  with:
    apiKey: ${{ secrets.HYPHEN_API_KEY }}
- uses: Hyphen/env-action@v1
  with:
    environment: production
    outputs: files
- id: deploy
  uses: Hyphen/deploy-action@v1
  with:
    environment: production
- run: echo "Deployed at ${{ steps.deploy.outputs.deployment-url }}"

Deploying Previews for Pull Requests

Pair the new Hyphen/deploy-action with Deployment Previews to give some or every pull request its own ephemeral, fully isolated environment: its own subdomain, its own infrastructure, no manual teardown.

Set the action's preview input to a name unique to the PR (e.g., its number) and Hyphen spins up a fresh preview off your existing deployment configuration. The workflow log links straight to the dashboard run, where reviewers can grab the preview URL and open it in a browser:

on:
  # Use pull_request to deploy a preview for every PR automatically,
  # workflow_dispatch to deploy on demand (e.g., for selected PRs only).
  pull_request:
    branches: [main]
  workflow_dispatch:

jobs:
  preview:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: Hyphen/setup-hx-action@v1
        with:
          apiKey: ${{ secrets.HYPHEN_API_KEY }}
      - id: deploy
        uses: Hyphen/deploy-action@v1
        with:
          preview: pr-${{ github.event.number }}
          prefix: pr-${{ github.event.number }}
      - run: echo "Preview run for PR #${{ github.event.number }}: ${{ steps.deploy.outputs.deployment-url }}"

Re-run the workflow on each push to update the preview. When the PR closes, delete the preview to tear down its infrastructure, or let Hyphen Agent's cleanup policies handle stale previews automatically.

Keep Reading