Back to Research

Cursor Gas Town: Multi-Agent Orchestration for Cursor CLI

We built Cursor Gas Town—a Cursor CLI implementation of Steve Yegge's Gas Town multi-agent orchestrator. Built in collaboration with the Cursor engineering team, it brings persistent work tracking and multi-agent coordination to Cursor workflows.

Key Takeaways

  • Cursor Gas Town adapts Gas Town's architecture for Cursor CLI workflows
  • Built in collaboration with the Cursor engineering team
  • Persistent work tracking survives agent crashes and restarts
  • Coordinate 20-30 agents across multiple projects with git-backed state
Rogier MullerJanuary 28, 20266 min read

Cursor Gas Town: Multi-Agent Orchestration for Cursor CLI

We've been working on something we're excited to share: Cursor Gas Town—a Cursor CLI implementation of Steve Yegge's Gas Town multi-agent orchestrator.

Built in collaboration with the Cursor engineering team, Cursor Gas Town brings persistent work tracking and multi-agent coordination to Cursor workflows.

What Is Gas Town?

Gas Town is a workspace manager that coordinates multiple AI coding agents working on different tasks. Instead of losing context when agents restart, Gas Town persists work state in git-backed hooks, enabling reliable multi-agent workflows.

The original Gas Town was built for Claude Code. We've adapted it for Cursor CLI, maintaining the same architecture while targeting the Cursor ecosystem.

The Core Problem It Solves

Challenge Gas Town Solution
Agents lose context on restart Work persists in git-backed hooks
Manual agent coordination Built-in mailboxes, identities, and handoffs
4-10 agents become chaotic Scale comfortably to 20-30 agents
Work state lost in agent memory Work state stored in Beads ledger

How It Works

Core Concepts

The Mayor — Your primary AI coordinator. The Mayor is a Cursor agent with full context about your workspace, projects, and agents. Start here—just tell the Mayor what you want to accomplish.

Town — Your workspace directory (e.g., ~/gt/). Contains all projects, agents, and configuration.

Rigs — Project containers. Each rig wraps a git repository and manages its associated agents.

Crew Members — Your personal workspace within a rig. Where you do hands-on work.

Polecats — Ephemeral worker agents that spawn, complete a task, and disappear.

Hooks — Git worktree-based persistent storage for agent work. Survives crashes and restarts.

Convoys — Work tracking units. Bundle multiple beads (tasks) that get assigned to agents.

Beads — Git-backed issue tracking system that stores work state as structured data.

The Workflow

  1. Create work as beads — Define tasks using the Beads system
  2. Group into convoys — Track related work together
  3. Assign to agents — Use gt sling to dispatch work
  4. Monitor progress — Track through convoy status
  5. Work persists — State survives agent restarts via git hooks

Installation

Prerequisites

  • Go 1.23+
  • Git 2.25+ (for worktree support)
  • beads (bd) 0.47.0+
  • tmux 3.0+ (recommended)
  • Cursor CLI

Setup

# Install Gas Town and Beads
go install github.com/cursorworkshop/cursor-gastown/cmd/gt@latest
go install github.com/steveyegge/beads/cmd/bd@latest
export PATH="$PATH:$HOME/go/bin"

# Create workspace
gt install ~/gt --git
cd ~/gt

# Add a project
gt rig add myproject https://github.com/you/repo.git

# Create crew workspace
gt crew add yourname --rig myproject

# Verify setup
gt doctor

Quick Start

# Navigate to your rig
cd ~/gt/myproject

# Create beads for your tasks
bd create "Add user authentication"
bd create "Fix login page styling"
bd create "Write integration tests"

# Create a convoy tracking your beads
gt convoy create "Auth Feature" gt-abc gt-def gt-ghi --notify

# Assign work to agents
gt sling gt-abc myproject

# Monitor progress
gt convoy list

Key Features

Workflow Durability

The most interesting aspect isn't the parallelism—it's how Gas Town handles durability.

Each agent has a "hook" where you assign work. If Cursor crashes or runs out of context, the workflow state lives in Git through the Beads system. A new agent picks up exactly where the last one stopped. No lost progress, no re-explaining context.

Multi-Agent Coordination

Gas Town coordinates multiple agents working on different tasks simultaneously. The Mayor agent dispatches work, and agents coordinate through git-backed hooks.

Persistent State

All work state is stored in git-backed hooks. This means:

  • Work survives agent crashes
  • You can resume from any point
  • Full history of changes
  • Easy rollback capability

Formulas

Gas Town includes built-in formulas for common workflows:

  • shiny — Full feature workflow (design → implement → review → test → submit)
  • code-review — Review existing code
  • security-audit — Security-focused review
  • design — Design-only phase
# Use a formula
gt sling shiny --var feature="User authentication" myproject

Built with Cursor Engineering

This project was built in collaboration with the Cursor engineering team. We worked closely with them to ensure Cursor Gas Town integrates seamlessly with Cursor CLI workflows and maintains compatibility with Cursor's agent architecture.

How This Relates to Our Methodology

At Cursor Workshop, we teach the Delegate, Review, Own framework. Cursor Gas Town fits this model perfectly:

Delegate: Gas Town handles the orchestration of boilerplate, standard implementations, and repetitive tasks across multiple agents. You describe what needs to happen; the agents figure out how.

Review: The Refinery role provides automated review, but humans still own the final verification. Gas Town produces work faster, which means more time for meaningful review.

Own: Strategic decisions—what to build, architectural choices, technology selection—remain human responsibilities. Gas Town amplifies execution but doesn't replace judgment.

Why We Built This

As we train teams on AI-assisted development, we've seen the limitations of single-agent workflows. Teams need to coordinate multiple agents working on different parts of a codebase simultaneously.

Gas Town solves this, but it was built for Claude Code. We wanted to bring the same capabilities to Cursor users. Working with the Cursor engineering team, we've adapted Gas Town's architecture for Cursor CLI.

Getting Started

Cursor Gas Town is open source and available on GitHub:

🔗 cursorworkshop/cursor-gastown

Built in collaboration with the Cursor engineering team, this is a Cursor CLI implementation of Steve Yegge's Gas Town multi-agent orchestrator.

For detailed documentation, installation guides, and examples, visit the repository.

The Future

This is just the beginning. We're continuing to work with the Cursor engineering team to improve multi-agent orchestration capabilities. As Cursor CLI evolves, so will Cursor Gas Town.

If you're using Cursor CLI and want to scale beyond single-agent workflows, Cursor Gas Town provides the infrastructure you need.


Cursor Gas Town is open source and available on GitHub. Built in collaboration with the Cursor engineering team. Contact us to learn how we can help your team adopt multi-agent workflows.

Want to learn more about Cursor?

We offer enterprise training and workshops to help your team become more productive with AI-assisted development.

Contact Us