Compoundry teaches you to build production multi-agent systems that actually run. Real architectures. Real coordination patterns. From someone running 10 agents daily, not someone who read a blog post about it.
Join The Build Sheet. One issue per week. No spam. Unsubscribe anytime.
The stuff nobody covers because they haven't built it yet.
How to structure systems where 5, 10, or 20 agents work together without stepping on each other. Routing, hierarchy, scope boundaries.
Handoff protocols, status tracking, escalation paths. How agents delegate work and report back without losing context or duplicating effort.
Short-term, long-term, shared memory patterns. How agents remember context, learn from past runs, and avoid repeating mistakes.
Taking agents from prototype to daily use. Error handling, failure recovery, monitoring, and the operational reality nobody talks about.
# Real config from a 10-agent household system # Not a demo. This runs every day. system: name: campbell-household coordinator: ivy escalation_target: gary agents: - name: ivy role: household-coordinator scope: routing, delegation, status-tracking can_delegate_to: [mason, hazel, cedar, brook] - name: mason role: home-maintenance scope: repairs, vendors, scheduling memory: short_term: current-tasks.md long_term: maintenance-history/ - name: hazel role: meal-planning scope: recipes, grocery-lists, dietary-prefs handoff_protocol: receives_from: ivy reports_to: ivy format: structured-status
Every architecture, pattern, and lesson comes from agents that are running in production right now. If it didn't work in practice, it doesn't get taught.
Gary runs a 10-agent household system, an AI consulting firm, and this platform. The curriculum is built from operational experience, not research papers.
The platform itself is built with agents. Content production, curriculum development, community management. We use what we teach.
Free. Weekly. Straight from the build log. No filler content, no recycled Twitter threads.
Your first 6 issues build a complete mental model for multi-agent systems. Each one stacks on the last.
I run a 10-agent household management system that handles everything from meal planning to home maintenance. I run an AI consulting firm that delivers with agent-assisted workflows. And I built this platform with the same agent architecture I teach.
I'm not an AI researcher. I'm not an influencer. I'm a builder who got tired of the gap between what people say AI agents can do and what actually works when you try to build them.
Compoundry exists because I needed this resource two years ago and it didn't exist. Everything here comes from building, breaking, fixing, and rebuilding real systems. That's the only way I know how to teach it.
One email per week. Real agent architectures, code you can use, lessons from production systems.
Free forever. No spam. Unsubscribe in one click.