I built dmx to help me write higher quality production code with lower token costs. dmx is a loop engineering framework that runs inside your Cursor or any other agentic IDE as an MCP server.

dmx comes with a set of built-in loops that orchestrates an AI SDLC. But, you can override them with YAML loop configs inside the repo. A dmx loop has trigger, skills, validators, memory and a goal state. And set success/ failure criteria and if/ how to chain to next loop.

You can try it by adding dmx into your IDE: { "mcpServers": { "dmx": { "command": "uvx", "args": ["--from", "deepmodel-dmx", "dmx", "serve"] } } }

The core project is open source. I'd love to get any feedback.

Suffers the same problem. Maybe it can be try in pi?