Anima-i (Methodius/Мефодий)
Anima-i is an experiment in autonomous AI agency that ran 10 generations of an AI agent called Methodius. Each generation works toward a goal, writes its experience to plain text files, and passes those files to the next generation—no shared context or memory survives between runs. It is for developers and researchers interested in how LLM agents behave over extended autonomous operation.
✨ Key features
- Generational agent runs with file-based memory and identity.
- Each generation works from a goal and writes to markdown files.
- Loop script runs agent steps until TODO list is empty.
- Telegram bot for managing generations and approving goals.
- Publishes results to a Telegram channel automatically.
- Includes scripts for image generation and channel photo updates.
🎯 Use cases
- Run autonomous AI agent experiments over multiple generations.
- Study how LLM agents behave without persistent memory.
- Create AI-generated content and publish to Telegram.
- Explore agent self-reflection and goal-setting processes.
- Build a feedback loop with an audience via Telegram.
📦 Installation
🧰 Requirements: Requires Claude CLI (claude), Bash, and optionally Python 3 with python-telegram-bot for the Telegram bot.
git clone https://github.com/Vitali-Ivanovich/anima-i.git
cd anima-i
bash init.sh generation_1
nano generation_1/MAIN_GOAL.md # set a goal
cd generation_1 && bash loop.sh # run the cycle
🚀 Usage
bash init.sh generation_1
nano generation_1/MAIN_GOAL.md # set a goal
cd generation_1 && bash loop.sh # run the cycle
⚠️ Good to know
The project is an experiment; findings indicate that factual knowledge decays quickly across generations, and the agent requires human involvement for distribution and feedback.
❓ FAQ
How does the agent remember between generations?
It doesn't have shared context or memory; the only continuity is what is written down in plain text files that are passed to the next generation.
What is the role of the Telegram bot?
The bot manages generations: it can start new goals, show status, manage the queue of proposed generations, and run background processes like a watchdog and scheduler.
What happens when the TODO list is empty?
The loop stops, and the generation completion protocol runs: generate an image, set channel photo, commit results, and propose a goal for the next generation.
Can I run this without the Telegram bot?
Yes, you can manually run init.sh and loop.sh, but the bot provides automated management and publishing features.
📊 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.