Viberglass Documentation

Viberglass routes bug reports and feature requests through AI coding agents. Tickets move through research, planning, and execution phases with human review gates between them. The agents run as clankers: workers that you provision on AWS Lambda, ECS Fargate, or Docker.

Quickstart — Clone the repo, docker compose up, and fix your first ticket in 10 minutes.

This documentation is split into two parts.

For Users

If you create tickets, review plans, run scheduled tasks, or use Viberglass through Slack, start here. These guides cover the web UI at app.viberglass.io and the Slack integration.

• Getting Started — Sign in, dashboard tour, and core concepts.
• Projects — Creating projects, linking repositories, and configuring auto-fix.
• Creating Tickets — File tickets from the web UI, Slack, or external integrations.
• Ticket Lifecycle — Research, planning, and execution phases with revision and approval flows.
• Claws (Scheduled Tasks) — Reusable task templates and recurring schedules.
• Secrets — Storing API keys and other secrets used by clankers.
• Integrations — GitHub, custom webhooks, and other ticketing/SCM connectors.
• Slack — Filing tickets and supervising agents from a Slack channel.

For Admins

If you operate Viberglass — deploy the platform, manage clanker container images, configure SSM secrets, or own the AWS infrastructure — start here.

• Platform Overview — How the pieces fit together.
• Architecture — Backend, worker, and frontend interactions.
• Platform Backend — Express API, PostgreSQL schema, and job orchestration.
• Viberator Worker — Agent harness, repo cloning, PR creation.
• Clankers — Agent types, deployment strategies, and configuration model.
• Clanker Images — Building and pushing worker container images to ECR.
• Deployment — Pulumi stacks, GitHub Actions, and Amplify.
• Secrets Management — SSM Parameter Store, KMS, and encryption.
• Troubleshooting — Common operational issues.