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.
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?
I haven't tried it. But, please do :) when you say same problem, token cost? or was it more about the code quality?