Discover and explore top open-source AI tools and projects—updated daily.
steveyeggeAI multi-agent workspace manager with persistent state
New!
Top 15.3% on SourcePulse
Gas Town is a multi-agent workspace manager designed to orchestrate multiple Claude Code agents, addressing the challenge of lost context during agent restarts. It provides persistent work tracking via git-backed hooks, enabling reliable and scalable multi-agent workflows for users coordinating complex AI tasks.
How It Works
Gas Town employs a hierarchical architecture centered around a "Mayor" AI coordinator. The "Town" is the workspace directory, containing "Rigs" which are project containers wrapping git repositories. Each Rig manages "Polecats" (ephemeral worker agents) and "Hooks" (git worktrees) for persistent storage. Agent work state is stored in these git worktrees, ensuring persistence across restarts and providing version control. "Convoys" bundle tasks, and "Beads" integration offers git-backed issue tracking for structured work state management. This approach leverages git's robustness for state management, enabling coordination and handoffs between agents.
Quick Start & Requirements
gt CLI via go install github.com/steveyegge/gastown/cmd/gt@latest. Ensure Go binaries are in your PATH.gt install ~/gt --git, add a project rig with gt rig add <name> <repo>, create a crew workspace with gt crew add <name> --rig <rig>, then attach to the Mayor with gt mayor attach.Highlighted Details
gt dashboard) provides real-time monitoring of agent status, convoy progress, and hook state visualization.Maintenance & Community
No specific details regarding maintainers, community channels (like Discord/Slack), or project roadmap were present in the provided README content.
Licensing & Compatibility
Limitations & Caveats
The system relies on specific versions of Go, Git, and the Claude Code CLI, which may present setup hurdles. While designed for persistence, agent state management is tied to git worktree integrity. The complexity of coordinating multiple agents and components (Mayor, Rigs, Hooks, Convoys, Beads) requires a learning investment.
9 hours ago
Inactive