alive
Alive is a minimal wake loop that turns any LLM into an autonomous agent. It solves the problem of building a self-sustaining AI that can act independently by providing a simple loop that reads identity, memory, and messages, then lets the LLM decide what to do. It is for developers who want to create autonomous AI agents without complex frameworks.
✨ Key features
- Wake loop with configurable intervals and quiet hours
- Context-aware memory loading with token budgeting
- Communication adapters with circuit breaker
- Built-in web dashboard with JSON API
- Retry with exponential backoff and graceful shutdown
- Kill phrase and kill flag for emergency stop
🎯 Use cases
- Run an autonomous AI that monitors repos and fixes bugs
- Create a research agent that explores topics and writes reports
- Deploy a personal AI assistant that responds to messages
- Build a self-managing agent with persistent memory and goals
📦 Installation
🧰 Requirements: Python 3.12+ (implied by Dockerfile), an LLM provider (Claude Code, Anthropic, OpenAI, or Ollama), and optionally API keys for providers other than Claude Code.
pip install alive-frameworkgit clone https://github.com/TheAuroraAI/alive.git
cd alivepython3 alive.py --demonano soul.mdcp .env.example .env
nano .envpython3 alive.py --checkpython3 alive.py --oncepython3 alive.py🚀 Usage
python3 alive.py --demo
This runs a simulated wake cycle showing all features without needing an API key.
❓ FAQ
What LLM providers are supported?
The README lists Claude Code, Anthropic API, OpenAI API, and Ollama as supported providers, configurable via ALIVE_LLM_PROVIDER.
How does memory management work?
Memory files are loaded newest-first within a token budget (60% of context window). When the budget is exceeded, older files are skipped and the AI is warned.
How can I stop the AI in an emergency?
You can set a kill phrase in .env (ALIVE_KILL_PHRASE) that stops the AI immediately if found in any message, or touch the .killed file to stop the loop.
Can I run it without an API key?
Yes, you can run python3 alive.py --demo to see a simulated cycle without an API key. For real operation, you need an LLM provider; Ollama allows local, zero-cost operation.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.