Docker Compose like process orchestrator for our agents dev environment
Docker Compose for local dev without containers; needs broader adoption beyond agent dev environments.
A CLI tool to orchestrate multi-repository Docker environments
Daemonizes foreground tasks (npm run dev) across repos, but solves a localized workflow problem.
Microservice developers, agency engineers managing multi-repo codebases
Docker Compose · Lerna · Turbo
I work at a software agency where I constantly have to deal with microservices and projects split across multiple repositories (e.g., Frontend, Backend API, Auth Service). Every morning, I found myself doing the same chore: opening multiple terminal tabs, navigating to different directories, and running `docker-compose up -d` or `npm run dev` one by one.
It was tedious, so I built *mdc (Multi-Docker-Compose)* to solve this exact problem.
*What it does:* `mdc` lets you define all your cross-repository commands in a single YAML file and orchestrates them with a single command: `mdc up`.
*Key features:* * *Parallel & Sequential Execution:* Boot up your entire stack concurrently, or sequentially if you have strict dependencies. * *Background Process Management:* It doesn't just run `docker-compose`. You can daemonize long-running foreground tasks like `npm run dev` (using `background: true` in the YAML) and easily monitor them with `mdc procs`. * *Clean UI:* Color-coded outputs and emojis to instantly see which project is failing or starting. * *Zero Dependencies:* Written in Go, distributed as a single binary. (This was actually my very first Go project!)
You can check it out here: https://github.com/tominaga-h/multi-docker-commander
I'd love to hear your feedback, thoughts on the architecture, or if you've faced similar local-dev nightmares. Happy to answer any questions!
Docker Compose for local dev without containers; needs broader adoption beyond agent dev environments.
Two-agent system where The Architect learns from PR review comments automatically.
Single Go binary: Telegram → Claude agents in isolated Docker with swarms, memory, Nix.
File-system driven agent config is clean, but OpenClaw and LangGraph already solve this.
Infinite canvas for agents beats terminal tab hell, but Steve Yegge's Beads tables are the real draw.
Example repo for YSA sandboxing when the main YSA product would be the actual submission.