Macross
A multi-agent platform, running in production.
-
Chat Room -
Message Tool -
Pipeline -
Roster
Click any image to view full size.
Macross is the multi-agent platform I use daily — rooms where specialized AI agents collaborate on actual work. It runs on Protoculture, an agentic harness with multi-provider LLM routing (Claude, Gemini, OpenAI), real-time orchestration, Docker-native deployment, and personality-driven agents.
- 3+ LLM providers routed
- 17 Live agents in rotation
- > 99.9% Uptime since deploy
Off-the-shelf agent frameworks were either vendor-locked to a single LLM, opinionated toward chatbot UIs, or fragile at multi-agent scale. Running more than two agents together in a shared context required custom orchestration.
Two layers. Protoculture — the message-driven harness — handles provider routing, turn-taking, interruption, memory, and tool access. Macross — the room-based app on top — is where agents collaborate on real work: rooms per project, personalities per agent, shared context, and a UI tuned for multi-party conversation instead of single-chatbot Q&A. Dockerized from day one.
Macross is the platform I use every day to run this portfolio build, sales outreach, QA reviews, and more. Protoculture — the harness underneath — has been running in production without regression for months.
- Node.js
- Docker
- Claude / Gemini / OpenAI
- WebSockets