📖Core ConceptsSentence → DNA → Execute

Sentence → DNA → Execute

Every Turtle follows a three-phase lifecycle: you say what you want, AI compiles it into code, and it runs forever.


Phase 1: Say What You Want

You describe your intent in plain language — one sentence is enough.

"Remind me to stand up every 25 minutes"

No configuration, no coding. Just describe the behavior you want.


Phase 2: AI Compiles into DNA

OpenTurtles AI reads your sentence and generates DNA — a piece of executable code that defines exactly how your Turtle should behave: what to observe, when to act, and how to respond.

~/.turtles/plans/pomodorostand.js

This compilation happens once. After DNA is generated, no AI is needed to run the Turtle. Zero ongoing LLM cost.

DNA is the blueprint of your Turtle. It captures your intent as deterministic logic — fast, predictable, and completely under your control.


Phase 3: Execute

Your Turtle runs locally as a background process, inside a sandboxed environment — isolated and secure. It can only access the tools and APIs you've explicitly connected.

  • Always running — starts on boot, survives reboots via the daemon
  • Lightweight — each Turtle uses minimal resources
  • Secure — sandboxed execution, no unauthorized access to your system
  • Offline capable — no internet required after compilation
  • Your Turtle watches the world, takes action when conditions are met, and keeps running — silently, reliably, forever.