MCPcopy Index your code
hub / github.com/aaif-goose/goosetown

github.com/aaif-goose/goosetown @main

Chat with this repo
repository ↗ · DeepWiki ↗ · + Follow
79 symbols 194 edges 10 files 7 documented · 9%
What it actually does AI analysis from the code graph — generated when you open this
loading…
README

Welcome to Goosetown!

Goosetown — a steampunk village of industrious geese

A tiny town that ships.

Quick Start · How It Works · AGENTS.md

Tests License Discord

Goosetown coordinates flocks of AI agents — researchers, writers, workers, reviewers — so you describe what to build and the town builds it. Research-first, parallel by default, with crossfire reviews across multiple models.

Quick Start

  1. Install goose (v1.25.0+)
  2. Clone and enter Goosetown bash git clone https://github.com/block/goosetown.git cd goosetown
  3. Start goose (the wrapper enables telepathy — orchestrator → delegate push messages) bash ./goose

Say hello — goose already knows its way around town.

[!NOTE] On first run, goose may ask you to set a few environment variables. Follow its instructions and restart.

See It Work

Here's what coordination looks like on the Town Wall:

[16:21:41] <orchestrator>       Spawning research flock...
[16:22:06] <researcher-local>   💡 Found existing patterns in GUIDES/
[16:22:19] <researcher-github>  🎬 Scanning issues and PRs
[16:23:46] <orchestrator>       Research complete. Dispatching workers...
[16:24:11] <worker-auth>        🎬 Claiming src/auth/mod.rs
[16:25:02] <reviewer-gpt5>     ✅ APPROVE (9/10)
  • The orchestrator spawns a research flock to understand the problem
  • Delegates broadcast progress on gtwall so siblings avoid duplicate work
  • The orchestrator synthesizes findings, dispatches workers, then runs crossfire review

How It Works

The orchestrator decomposes your request into phases — research, build, review — and dispatches parallel delegates called flocks that coordinate through the Town Wall.

        Orchestrator
            │ spawns
     ┌──────┼──────┐
     ▼      ▼      ▼
 Researchers (flock)      ← share findings via gtwall
            │ synthesize
     ┌──────┼──────┐
     ▼      ▼      ▼
 Workers + Writers        ← parallel execution
            │ review
     ┌──────┼──────┐
     ▼      ▼      ▼
 Reviewers (crossfire)    ← multi-model adversarial QA
            │
     Final deliverable

When three or more delegates share a task and coordinate via gtwall, that's a flock.

Core Concepts

  • Orchestrator — the main session; decomposes work, dispatches delegates, synthesizes results
  • Delegates — parallel subagents: researchers, writers, workers, reviewers
  • Skills — role definitions loaded into each delegate at spawn
  • gtwall — the Town Wall; broadcast channel for real-time delegate coordination
  • Telepathy — orchestrator → delegate push messages for urgent paging

There's a real-time dashboard for watching your flock work (yes, they're actual geese on a map) — just ask goose to launch it.

Goosetown Village Dashboard — real-time agent coordination view

Learn more in AGENTS.md.

Acknowledgments

Goosetown was inspired by Gas Town, the multi-agent coordination framework created by Steve Yegge. His blog post announcing Gas Town laid out the vision of orchestrating flocks of AI agents — researchers, workers, reviewers — that Goosetown builds on.


Part of the goose ecosystem by Block.

Apache 2.0 License

Core symbols most depended-on inside this repo

Shape

Function 71
Method 5
Class 3

Languages

TypeScript87%
Python13%

Modules by API surface

ui/js/components.js25 symbols
ui/js/app.js14 symbols
ui/js/village.js12 symbols
ui/js/editor.js11 symbols
tests/test_catalog.py10 symbols
ui/js/state.js4 symbols
ui/js/tiles.js2 symbols
ui/js/buildings.js1 symbols

For agents

$ claude mcp add goosetown \
  -- python -m otcore.mcp_server <graph>

⬇ download graph artifact

Ask about this repo answers extend the page