Skip to content

Gas Town

Coordinates multiple AI coding agents across projects, saving their work state in git so it survives restarts.

Stars
18.7k
Forks
1.7k
Licence
MIT
Language
Go
Last update
9 days ago
Type
Dev tool

What it is

Gas Town is a workspace manager that lets you run several AI coding agents, such as Claude Code, GitHub Copilot, Codex, and Gemini, on different tasks at once. It stores work state in git-backed hooks and a Beads issue ledger, so agents can pick up where they left off after a restart. A coordinator agent called the Mayor is the main entry point for giving instructions.

  • Hand off a set of coding tasks to a coordinator agent and let worker agents execute them in parallel
  • Track multi-step work as bead-based issues bundled into convoys
  • Query earlier agent sessions for context using the seance command
  • Route blocked work up through escalation levels using gt escalate

Great for

  • Developers already using Claude Code or similar AI coding agents on several tasks at once
  • Teams who want agent work tracked as git-backed issues
  • People running many agents in parallel who need restart-safe state
  • Users comfortable with a Go toolchain and a terminal-based workflow

Not ideal for

  • People who only use one AI agent occasionally, since the coordination layer adds setup overhead
  • Users without git, tmux, or comfort with command-line tools, which the install path requires
  • Anyone needing a polished graphical app, as the README describes a CLI and tmux-based workflow
Step-by-step guide: how to use Gas TownInstall, first run, examples, tips and FAQs

How to get it

  1. 1

    Install with a package manager

    You'll need: Git 2.20+ · Go 1.26.2+ (Linux, Windows, and macOS source builds) · Beads (bd) 0.57.0+ for native installs · tmux 3.0+ for gt up · Dolt (required on Linux and Windows)

    Install on macOS with Homebrew

    brew install gastown

    Install gt on Linux

    go install github.com/steveyegge/gastown/cmd/gt@latest

    Install bd on Linux

    go install github.com/steveyegge/beads/cmd/bd@latest
  2. 2

    Run it

    • Install the prerequisites listed in the README, including tmux and a runtime such as Claude Code
    • Install gt and bd for your platform using the commands above
    • Create a town workspace, for example at ~/gt/, following the README setup
    • Start the system with gt up
    • Tell the Mayor what you want to accomplish
  3. 3

    Get the most out of it

    • Use convoys with the mountain label for epic-scale work that needs autonomous stall detection
    • Set scheduler.max_polecats to enable deferred dispatch and limit API rate limit exhaustion
    • Read the glossary and linked design docs for molecules, escalation, and the scheduler
    • Use gt seance to discover and question predecessor agent sessions

Downloads

Strengths

  • Persists work state in git-backed hooks so agent progress survives crashes and restarts
  • Supports several AI runtimes, including Claude Code, Copilot, Codex, and Gemini, according to the README
  • Includes built-in monitoring and a merge queue that runs verification before merging to main

Trade-offs

  • Setup has many prerequisites (Git, Go, Beads, Dolt, tmux, ICU4C on source builds) and platform-specific steps
  • The README warns against go install on macOS due to Gatekeeper, so install paths differ by platform
  • Some features, like the Wasteland federation and mountain convoys, are described only briefly in the truncated README

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.