Overview

Gas Town is an advanced multi-agent workspace manager designed to orchestrate complex workflows involving AI coding assistants like Claude Code, GitHub Copilot, and others. It provides persistent work tracking and reliable context management, solving the common problem of AI agents losing state upon restart. Gas Town enables scaling from a few agents to dozens, transforming chaotic development environments into highly organized, productive systems.

Key Features

• Persistent Work State: Utilizes git-backed hooks to store and recover agent work, ensuring continuity even after crashes or restarts.
• Multi-Agent Coordination: Manages identities, mailboxes, and seamless handoffs between multiple AI agents for sophisticated task delegation.
• Scalable Architecture: Designed to comfortably manage 20-30 agents, with built-in monitoring and health systems (Witness, Deacon, Dogs) to maintain stability.
• GitOps Integrated: Leverages git worktrees for persistent storage and a Bors-style merge queue (Refinery) for safe, automated code integration.
• Work Tracking (Beads & Convoys): Employs a git-backed issue tracking system (Beads) for structured work items and Convoys for bundling and tracking progress of multiple tasks.
• Workflow Automation (Molecules): Supports TOML-defined workflow templates (Molecules) for automating multi-step processes.
• Context Recovery (Seance): Enables agents to query their predecessors for historical context and decisions, preventing redundant work.
• Federated Work (Wasteland): Allows connecting multiple Gas Towns into a network for distributed work claiming and contribution.
• Interactive Monitoring (gt feed): A terminal UI for real-time activity, agent health, and convoy progress, with a dedicated "problems" view for identifying and resolving agent issues.
• Web Dashboard: Provides a centralized, auto-refreshing view of workspace status, agents, convoys, and issue tracking, accessible via a browser.

Typical Use Cases

• Complex Feature Development: Orchestrate multiple AI agents to tackle large features, breaking them down into smaller, trackable "beads" managed within convoys.
• Bug Triage and Resolution: Automate the process of identifying, assigning, and fixing bugs with agents that can escalate blockers and report completion.
• Code Refactoring at Scale: Coordinate agents for large-scale code refactors, ensuring context is maintained throughout the process.
• Automated Release Management: Utilize built-in or custom "formulas" to define and execute repeatable processes like version bumping, testing, and publishing.
• CI/CD Pipeline Integration: Leverage the Refinery's merge queue to ensure only verified code is merged into main branches, integrating seamlessly with development workflows.
• Research and Prototyping: Rapidly spin up AI agents to explore different solutions or prototype new functionalities, with persistent state allowing for iterative development.