GitHub Actions

Run Station AI agents directly in your GitHub workflows without installing Station locally. The station-action supports three modes for loading agents.

Quick Start

The easiest way to add a Station agent workflow to your repository is using the CLI:

stn github init

This interactive command will:

Ask how you want to provide agents (CloudShip bundle ID or local environment)
Help you select an agent to run
Configure the workflow trigger (push, PR, schedule, manual)
Generate .github/workflows/station-agent.yml
Tell you which GitHub secrets to configure

Example session:

$ stn github init

🚀 Station GitHub Actions Setup
================================

How would you like to provide agents?
  1. CloudShip Bundle ID (recommended - keeps repo clean)
  2. Local environment (agents stored in repo)

Choice [1]: 1
Enter your CloudShip Bundle ID: 970fa728-4276-4443-9e51-35919b650d7a
Enter agent name to run: posthog-dashboard-reporter

When should this workflow run?
  1. On push to main branch
  2. On pull requests
  3. On a schedule (e.g., daily)
  4. Manual trigger only

Choice [4]: 4

✅ GitHub Actions workflow created!

📄 File: .github/workflows/station-agent.yml

🔐 Required GitHub Secrets:
   OPENAI_API_KEY      - Your OpenAI API key
   CLOUDSHIP_API_KEY   - Your CloudShip API key

Set secrets at: https://github.com/<owner>/<repo>/settings/secrets/actions

Manual Setup

Alternatively, you can manually create the workflow. The simplest approach is using a CloudShip bundle ID:

name: Run AI Agent

on: [push]

jobs:
  analyze:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - uses: cloudshipai/station-action@main
        with:
          agent: 'posthog-dashboard-reporter'
          task: 'List all my dashboards and summarize my analytics setup.'
          bundle-id: '970fa728-4276-4443-9e51-35919b650d7a'
          cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
          POSTHOG_API_KEY: ${{ secrets.POSTHOG_API_KEY }}

Agent Source Modes

Station action supports three ways to load agents:

Bundle ID (Recommended)
Bundle URL
Local Environment

Use a bundle published to CloudShip. This is the cleanest approach - no Station files need to be committed to your repository.Step 1: Publish your bundle to CloudShipFrom your local Station workspace:

# Share an environment as a bundle
stn bundle share default

# Output:
# Bundle shared successfully!
# Bundle ID: 970fa728-4276-4443-9e51-35919b650d7a

Step 2: Use the bundle ID in your workflow

- uses: cloudshipai/station-action@main
  with:
    agent: 'my-agent'
    task: 'Perform the task'
    bundle-id: '970fa728-4276-4443-9e51-35919b650d7a'
    cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
  env:
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Get your CloudShip API key from app.cloudshipai.com under Settings > API Keys.

Download a bundle from any URL (S3, GitHub Releases, etc.):

- uses: cloudshipai/station-action@main
  with:
    agent: 'my-agent'
    task: 'Perform the task'
    bundle-url: 'https://github.com/myorg/myrepo/releases/download/v1.0.0/bundle.tar.gz'
  env:
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Creating a bundle for URL distribution:

# Build a bundle locally
stn bundle build default -o ./my-bundle.tar.gz

# Upload to your release/storage
gh release upload v1.0.0 ./my-bundle.tar.gz

Use agents defined in your repository’s environments/ folder:

- uses: cloudshipai/station-action@main
  with:
    agent: 'my-agent'
    task: 'Perform the task'
    environment: 'default'
  env:
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Repository structure:

your-repo/
├── environments/
│   └── default/
│       ├── agents/
│       │   └── my-agent.yaml
│       └── mcp/
│           └── my-tools.yaml
└── .github/
    └── workflows/
        └── agent.yml

This mode is useful when you want to version control your agents alongside your application code.

Action Inputs

Input	Description	Required	Default
`agent`	Agent name to run	Yes	-
`task`	Task description for the agent	Yes	-
`bundle-id`	CloudShip bundle ID	No*	-
`bundle-url`	URL to download bundle from	No*	-
`environment`	Local environment name	No*	`default`
`cloudship-api-key`	CloudShip API key (for bundle-id mode)	No	-
`provider`	AI provider (`openai`, `anthropic`, `gemini`, `ollama`)	No	`openai`
`model`	Model override	No	Provider default
`base-url`	Custom API endpoint	No	-
`timeout`	Execution timeout in seconds	No	`300`
`max-steps`	Maximum agent steps	No	`50`

*At least one of bundle-id, bundle-url, or environment must be specified. If none are provided, the action looks for a local environments/default folder.

AI Provider Configuration

OpenAI
Anthropic
Google Gemini

- uses: cloudshipai/station-action@main
  with:
    agent: 'my-agent'
    task: 'Analyze the codebase'
    bundle-id: 'your-bundle-id'
    cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
    provider: 'openai'
    model: 'gpt-4o'
  env:
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

- uses: cloudshipai/station-action@main
  with:
    agent: 'my-agent'
    task: 'Analyze the codebase'
    bundle-id: 'your-bundle-id'
    cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
    provider: 'anthropic'
    model: 'claude-sonnet-4-20250514'
  env:
    ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}

- uses: cloudshipai/station-action@main
  with:
    agent: 'my-agent'
    task: 'Analyze the codebase'
    bundle-id: 'your-bundle-id'
    cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
    provider: 'gemini'
    model: 'gemini-2.0-flash-exp'
  env:
    GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}

Complete Examples

Analytics Reporter

Run a PostHog analytics agent to summarize your dashboards:

name: Weekly Analytics Report

on:
  schedule:
    - cron: '0 9 * * 1'  # Monday 9 AM
  workflow_dispatch:

jobs:
  report:
    runs-on: ubuntu-latest
    steps:
      - uses: cloudshipai/station-action@main
        with:
          agent: 'posthog-dashboard-reporter'
          task: |
            Generate a weekly analytics report:
            1. List all dashboards
            2. Summarize key metrics from each
            3. Highlight any significant changes
          bundle-id: '970fa728-4276-4443-9e51-35919b650d7a'
          cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
          POSTHOG_API_KEY: ${{ secrets.POSTHOG_API_KEY }}

PR Code Review

Review pull requests automatically:

name: AI Code Review

on:
  pull_request:
    types: [opened, synchronize]

jobs:
  review:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: write

    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - uses: cloudshipai/station-action@main
        with:
          agent: 'code-reviewer'
          task: |
            Review the changes in this PR. Focus on:
            - Security vulnerabilities
            - Performance issues
            - Code quality and best practices
            - Potential bugs
          bundle-id: 'your-code-review-bundle-id'
          cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Security Scan on Push

Scan for security issues on every push:

name: Security Scan

on:
  push:
    branches: [main]

jobs:
  scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - uses: cloudshipai/station-action@main
        with:
          agent: 'security-scanner'
          task: 'Scan the codebase for security vulnerabilities, exposed secrets, and dependency issues'
          bundle-id: 'your-security-bundle-id'
          cloudship-api-key: ${{ secrets.CLOUDSHIP_API_KEY }}
          timeout: '600'
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Publishing Bundles to CloudShip

To use bundle-id mode, first publish your agents to CloudShip:

# 1. Create agents in your local Station workspace
stn init
stn agent add --name "my-agent" --description "Does something useful"

# 2. Add MCP tools
stn mcp add --name "my-tools" --url "https://api.example.com/mcp"

# 3. Test locally
stn run "my-agent" --task "Test the agent"

# 4. Share to CloudShip
stn bundle share default

# Output shows your bundle ID:
# Bundle ID: abc123-def456-...

Store your bundle ID somewhere safe. You’ll need it for the bundle-id input in your workflows.

Build-Time Actions

Station also provides reusable actions for building bundles and Docker images.

Build Bundle

Create distributable bundles from environments:

- uses: cloudshipai/station/.github/actions/build-bundle@main
  id: bundle
  with:
    environment: 'production'
    version: '1.0.0'

- run: echo "Bundle created at ${{ steps.bundle.outputs.bundle-path }}"

Build Docker Image

Build and push Docker images:

- uses: cloudshipai/station/.github/actions/build-image@main
  with:
    source-type: 'environment'
    environment: 'production'
    image-name: 'my-org/station-production'
    image-tag: 'latest'
    push: 'true'
    registry-username: ${{ github.actor }}
    registry-password: ${{ secrets.GITHUB_TOKEN }}

Setup Station CLI

Install Station CLI without running an agent:

- uses: cloudshipai/station/.github/actions/setup-station@main
  with:
    version: 'latest'
  env:
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

- run: stn --version

Best Practices

Use bundle-id mode for production

Bundle-id mode keeps your repository clean - no Station configuration files to maintain. Agents are versioned separately in CloudShip.

Store API keys as GitHub secrets

Never hardcode API keys. Use GitHub secrets:

env:
  OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
  CLOUDSHIP_API_KEY: ${{ secrets.CLOUDSHIP_API_KEY }}

Set appropriate timeouts

Complex tasks may need longer timeouts:

with:
  timeout: '600'  # 10 minutes

Use workflow_dispatch for testing

Add manual trigger for easy testing:

on:
  workflow_dispatch:
  push:
    branches: [main]

Troubleshooting

Agent not found

Ensure the agent name matches exactly (case-sensitive). List available agents:

stn agent list

Bundle download failed

For bundle-id mode, verify:

cloudship-api-key is set correctly
Bundle ID is valid and not deleted
Your CloudShip account has access to the bundle

MCP tools not available

Ensure required environment variables are passed:

env:
  POSTHOG_API_KEY: ${{ secrets.POSTHOG_API_KEY }}
  DATABASE_URL: ${{ secrets.DATABASE_URL }}

Next Steps

Bundles - Learn more about packaging agents
GitOps Workflow - Version control your agents
Docker Deployment - Container deployment options

Get Started

Station CLI

Building Agents

Station Lattice

Deployment

Advanced

GitHub Actions

GitHub Actions

Quick Start

Manual Setup

Agent Source Modes

Action Inputs

AI Provider Configuration

Complete Examples

Analytics Reporter

PR Code Review

Security Scan on Push

Publishing Bundles to CloudShip

Build-Time Actions

Build Bundle

Build Docker Image

Setup Station CLI

Best Practices

Troubleshooting

Next Steps

​GitHub Actions

​Quick Start

​Manual Setup

​Agent Source Modes

​Action Inputs

​AI Provider Configuration

​Complete Examples

​Analytics Reporter

​PR Code Review

​Security Scan on Push

​Publishing Bundles to CloudShip

​Build-Time Actions

​Build Bundle

​Build Docker Image

​Setup Station CLI

​Best Practices

​Troubleshooting

​Next Steps

GitHub Actions

Quick Start

Manual Setup

Agent Source Modes

Action Inputs

AI Provider Configuration

Complete Examples

Analytics Reporter

PR Code Review

Security Scan on Push

Publishing Bundles to CloudShip

Build-Time Actions

Build Bundle

Build Docker Image

Setup Station CLI

Best Practices

Troubleshooting

Next Steps