gastown — What is it?

Gas Town is a multi-agent workspace manager designed to orchestrate and coordinate AI coding agents like Claude Code and GitHub Copilot, ensuring persistent work tracking and reliable workflows.

⭐ 16,978 Stars 🍴 1,556 Forks Go MIT Author: gastownhall
Source: README View on GitHub →

Why it matters

Gas Town is gaining attention due to its innovative approach to managing AI coding agents, addressing the challenges of context loss and manual coordination. Its unique use of git-backed hooks for persistent work state and its ability to scale to a large number of agents are notable technical choices.

Source: README

Core Features

Multi-agent orchestration

Coordinates multiple AI coding agents, such as Claude Code and GitHub Copilot, allowing them to work on different tasks without losing context.

Source: README
Git-backed hooks

Utilizes git worktree-based persistent storage for agent work, ensuring that work state survives crashes and restarts.

Source: README
Rig architecture

Project containers that wrap git repositories and manage associated agents, providing a structured environment for each project.

Source: README
Beads integration

A Git-backed issue tracking system that stores work state as structured data, enabling efficient work item management.

Source: README
Molecules

Workflow templates that coordinate multi-step work, with options for lightweight wisps and sub-wisps with checkpoint recovery.

Source: README
Monitoring system

A three-tier watchdog system (Witness, Deacon, Dogs) to keep agents healthy, monitor lifecycle, detect stuck agents, and manage session cleanup.

Source: README
Refinery

Per-rig merge queue processor that batches merge requests, runs verification gates, and merges to main using a bisecting queue.

Source: README
Escalation system

Severity-routed issue escalation that routes blocked agents through a Deacon, Mayor, and Overseer for resolution.

Source: README
Scheduler

Config-driven capacity governor for polecat dispatch, preventing API rate limit exhaustion by batching dispatch under configurable concurrency limits.

Source: README
Seance

Session discovery and continuation, allowing agents to query previous sessions for context and decisions from earlier work.

Source: README
Wasteland

A federated work coordination network linking Gas Towns through DoltHub, enabling rigs to post wanted items, claim work from other towns, and earn portable reputation.

Source: README

Architecture

Gas Town follows a modular architecture with distinct components such as the Mayor (AI coordinator), Town (workspace directory), Rigs (project containers), Crew Members (personal workspaces), Polecats (worker agents), Hooks (persistent storage), and Convoys (work tracking units). The system leverages git for version control and persistent storage, with a focus on scalability and reliability.

Source: README

Project Knowledge Graph

Knowledge graph: project (center) + core features (inner hexagons) + key dependencies (outer chips) github.com/steveyegge/beadsgithub.com/ste… github.com/steveyegge/gastowngithub.com/ste… github.com/dolthub/doltgithub.com/dol… github.com/go-rod/rodgithub.com/go-… go.opentelemetry.io/otelgo.opentelemet… Multi-agent orchestrationMulti-agent orchest… Git-backed hooks Rig architecture Beads integration Molecules Monitoring system Refinery Escalation system gastown Project Core feature Key dependency

Center: project; inner ring: core feature modules; outer ring: key dependencies. Auto-generated from core_features and tech_stack.key_deps.

Tech Stack

LanguageGoFrameworkCobra, Beads, Dolt, Rod, Otel, Lumberjack, Toml, YAML, ICU4C, and others
github.com/steveyegge/beadsgithub.com/steveyegge/gastowngithub.com/dolthub/doltgithub.com/go-rod/rodgo.opentelemetry.io/otel
Docker Compose
Source: Dependency files + code tree

Quick Start

Install the required tools (Git, Go, Beads, ICU4C, tmux, Claude Code CLI). Create your workspace with `gt install ~/gt --shell --git`. Start the long-lived services with `gt up`.
Source: README Installation/Quick Start

Use Cases

Gas Town is suitable for developers and teams working with multiple AI coding agents, especially those requiring persistent work tracking and reliable workflows. It is useful for managing complex projects with a large number of agents, ensuring efficient coordination and task completion.

Source: README

Strengths & Limitations

Strengths

  • Strength 1: Innovative approach to managing AI coding agents with persistent work tracking.
  • Strength 2: Scalable architecture that can handle a large number of agents.
  • Strength 3: Robust monitoring and escalation systems to ensure agent health and task completion.

Limitations

  • Limitation 1: Requires a significant setup and configuration process.
  • Limitation 2: May have a steep learning curve for new users.
Source: Synthesis of README, code structure and dependencies

Latest Release

v1.2.1 (2026-06-06): Pre-compiled binaries for Linux, macOS (Intel & Apple Silicon), and Windows.

Source: GitHub Releases

Verdict

Gas Town is a promising project for teams looking to streamline the management of AI coding agents in complex development workflows. Its innovative architecture and focus on reliability make it a valuable tool for organizations requiring robust and scalable solutions.

Frequently Asked Questions

What is gastown?

Gas Town is a multi-agent workspace manager designed to orchestrate and coordinate AI coding agents like Claude Code and GitHub Copilot, ensuring persistent work tracking and reliable workflows.

What are the main features of gastown?

gastown's core features include: Multi-agent orchestration, Git-backed hooks, Rig architecture, Beads integration, Molecules.

Why is gastown trending?

Gas Town is gaining attention due to its innovative approach to managing AI coding agents, addressing the challenges of context loss and manual coordination.

What is gastown used for?

Gas Town is suitable for developers and teams working with multiple AI coding agents, especially those requiring persistent work tracking and reliable workflows.

Transparency Notice
This page is auto-generated by AI (a large language model) from the following public materials: GitHub README, code tree, dependency files and release notes. Analyzed at: 2026-07-10 18:32. Quality score: 85/100.

Data sources: README, GitHub API, dependency files