I was spending more time orchestrating Claude Code and Cursor than actually coding. Run command → wait → check output → repeat.
So I built v16: persistent AI agents that work autonomously on my laptop.
- Each agent: ~40MB Go process
- Chat via Telegram (@devops, @research, @monitor)
- Run cron jobs autonomously (git checks, research, monitoring)
- Multi-LLM support (Claude, GPT-4, Groq)
My MacBook runs 4 agents 24/7 at ~160MB RAM. They handle git commits, compile research, and alert on system issues.
Written in Go. Persistent memory via JSON. Battery-aware.
Open source: https://github.com/anup-singhai/v16
Blog: https://v16.ai/blog/army-of-ai-agents