Gastown
A Git-based coordination tool for multiple AI coding agents that provides a persistent orchestration layer for coding agents such as Claude Code and GitHub Copilot. It manages work state through structured hooks, enabling concurrent collaboration and task tracking across dozens of agents simultaneously.
Timeline (last 90 days)
Beads, from the Gastown ecosystem, replaces an agent's draft notes with a structured issue ledger, writing each state change as an issue with an ID, status, and parent-child relationships.
Gastown stores WorkState in Git-backed structured hooks rather than relying on the Agent's own memory.
Gastown is suited for scenarios that need to scale to dozens of Agents working simultaneously, reliably track complex multi-step tasks, and require workflows to survive crashes or restarts.
For scenarios where a single Agent completes simple tasks, Gastown constitutes over-engineering.
Gastown organizes workflows using a town metaphor, with the Mayor serving as the primary coordinator responsible for orchestrating task scheduling across the Agent team.
Gastown is a Git-based, persistent workspace manager that provides an orchestration layer for coding agents such as Claude Code and GitHub Copilot.
Once configured, Gastown automatically dispatches Agents to their respective tasks and maintains process visibility through a real-time system.
Through a separation-of-concerns design that decouples orchestration logic from execution logic, swapping out the underlying Agent theoretically does not affect the operation of the entire workflow.
Writing WorkState to Git gives the Agent collaboration history an auditable property, enabling rollback to any historical node and using diffs to locate the Agent and step that introduced a problem.
The material in this article comes from a brief product introduction video; deeper performance characteristics and real-world implementation results still require further verification against official documentation and actual usage.
1 more timeline events
All Facts (11)
Beads, from the Gastown ecosystem, replaces an agent's draft notes with a structured issue ledger, writing each state change as an issue with an ID, status, and parent-child relationships.
50%UnverifiedGastown stores WorkState in Git-backed structured hooks rather than relying on the Agent's own memory.
50%UnverifiedGastown is suited for scenarios that need to scale to dozens of Agents working simultaneously, reliably track complex multi-step tasks, and require workflows to survive crashes or restarts.
50%UnverifiedFor scenarios where a single Agent completes simple tasks, Gastown constitutes over-engineering.
50%UnverifiedGastown organizes workflows using a town metaphor, with the Mayor serving as the primary coordinator responsible for orchestrating task scheduling across the Agent team.
50%UnverifiedGastown is a Git-based, persistent workspace manager that provides an orchestration layer for coding agents such as Claude Code and GitHub Copilot.
50%UnverifiedOnce configured, Gastown automatically dispatches Agents to their respective tasks and maintains process visibility through a real-time system.
50%UnverifiedThrough a separation-of-concerns design that decouples orchestration logic from execution logic, swapping out the underlying Agent theoretically does not affect the operation of the entire workflow.
50%UnverifiedWriting WorkState to Git gives the Agent collaboration history an auditable property, enabling rollback to any historical node and using diffs to locate the Agent and step that introduced a problem.
50%UnverifiedThe material in this article comes from a brief product introduction video; deeper performance characteristics and real-world implementation results still require further verification against official documentation and actual usage.
50%UnverifiedIn Gastown, rigs correspond to codebases or units of work, and convoys are used to group and manage tasks.
50%