Auggie Agent GitHub Action

AI-powered code assistance for GitHub pull requests using Auggie.

Quick Start

1. Get Your Augment API Credentials

First, you'll need to obtain your Augment Authentication information from your local Auggie session:

Example session JSON:

{
  "accessToken": "your-api-token-here",
  "tenantURL": "https://your-tenant.api.augmentcode.com"
}

There are 2 ways to get the credentials:

• Run auggie --print-augment-token
  • Copy the JSON after TOKEN=
• Copy the credentials stored in your Augment cache directory, defaulting to ~/.augment/session.json

⚠️ Security Warning: These tokens are OAuth tokens tied to your personal Augment account and provide access to your Augment services. They are not tied to a team or enterprise. Treat them as sensitive credentials:

• Never commit them to version control
• Only store them in secure locations (like GitHub secrets)
• Don't share them in plain text or expose them in logs
• If a token is compromised, immediately revoke it using auggie --revoke-all-augment-tokens

2. Set Up the GitHub Repository Secret

You need to add your Augment credentials to your GitHub repository:

Adding Secret

1. Navigate to your repository on GitHub
2. Go to Settings → Secrets and variables → Actions
3. Add the following:
   • Secret: Click "New repository secret"
     • Name: AUGMENT_SESSION_AUTH
     • Value: The json value from step 1

Need more help? For detailed instructions on managing GitHub secrets, see GitHub's official documentation:

• Using secrets in GitHub Actions

3. Create Your Workflow File

Add a new workflow file to your repository's .github/workflows/ directory and merge it.

Example Workflows

For complete workflow examples, see the example-workflows/ directory which contains:

• PR Description Generation - Automatically generate comprehensive PR descriptions
• Code Review - Perform automated code quality and security reviews
• Issue Triage - Automatically analyze and triage GitHub issues with priority and categorization recommendations
• Template-Based Review - Demonstrates the template system for dynamic, context-aware instructions

Each example includes a complete workflow file that you can copy to your .github/workflows/ directory and customize for your needs.

Advanced

Inputs

Input	Description	Required	Example
augment_session_auth	Augment session authentication JSON (store as secret)	No**	${{ secrets.AUGMENT_SESSION_AUTH }}
augment_api_token	API token for Augment services (store as secret)	No**	${{ secrets.AUGMENT_API_TOKEN }}
augment_api_url	Augment API endpoint URL (store as variable)	No**	${{ vars.AUGMENT_API_URL }}
github_token	GitHub token with repo and user:email scopes.	No	${{ secrets.GITHUB_TOKEN }}
instruction	Direct instruction text for simple commands	No*	"Generate PR description"
instruction_file	Path to file with detailed instructions	No*	/tmp/instruction.txt
template_directory	Path to template directory for dynamic instructions	No*	.github/templates
template_name	Template file name (default: prompt.njk)	No	pr-review.njk
pull_number	PR number for template context extraction	No	${{ github.event.pull_request.number }}
repo_name	Repository name for template context	No	${{ github.repository }}
custom_context	Additional JSON context for templates	No	'{"priority": "high"}'
model	Model to use; passed through to auggie as --model	No	e.g. sonnet4, from auggie --list-models

*Either instruction, instruction_file, or template_directory must be provided.

**Either augment_session_auth OR both augment_api_token and augment_api_url must be provided for authentication.

Template System

For advanced use cases, the Auggie Agent supports a template system that automatically extracts context from GitHub pull requests and allows you to create dynamic, reusable instruction templates. Templates are ideal when you need instructions that adapt based on PR content, file changes, or custom data.

See TEMPLATE.md for complete documentation on creating and using templates.