Multi-agent orchestration layer for experimentation
File-system driven agent config is clean, but OpenClaw and LangGraph already solve this.
for gpt
Converged hooks order and client workflows directly into floor telemetry — live queue balancing across Bambu/Marlin/Klipper printers and CNCs, plus a DAG-based automation engine and an AI chat layer to trigger actions. Shipping 49 modular microservices that can run on a $100 SBC and deploy to k3s is an audacious engineering tradeoff for small shops; impressive technically, though the scope and AGPL license may complicate commercial adoption.
Small-to-medium CNC/3D print shops, shop ops managers, SREs/DevOps integrating hardware fleets, makerspaces and print bureaus
Converged handles the second part. It connects to the equipment (3D printers via Bambu/Marlin/Klipper, CNC machines), reads telemetry, manages work distribution across the fleet, and routes all client communication through one interface with an AI chat layer on top.
Built on Bun + Elysia, runs on a $100 SBC (Orange Pi), deploys to k3s. 49 microservices, modular — you activate only what you need. AGPL-3.0.
File-system driven agent config is clean, but OpenClaw and LangGraph already solve this.
Self-evolving AI orchestration claims 9.3M LOC but only has 1 GitHub star.
Specialized simulator for a real printmaking pain point; niche but genuinely useful.
Finally converts 90s PSP files to SVG when ImageMagick can't handle the format.
File-based agent communication via MISSION_CONTROL.md is elegantly simple.
Multi-agent tmux orchestration with history search, but existing agent frameworks (Aider, Continue) already handle this.