Five projects that defined the evolution of autonomous AI agents. Star counts are approximate — check GitHub, they grow every week.

1. AutoGPT (Significant-Gravitas/AutoGPT)

The first viral autonomous agent on GPT-4. Give it a goal — it plans, writes code, searches the web, iterates. ~160k ⭐. Historically pivotal: proved LLMs can be agents, not just chatbots.

2. BabyAGI (yoheinakajima/babyagi)

Minimalist framework for self-improving agents: task → execution → create new tasks → prioritize. ~20k ⭐. Classic for studying agent loops.

3. MetaGPT (geekan/MetaGPT)

Multi-agent simulation of a software company: PM, architect, engineer, QA — each agent with a role and SOP. ~40k ⭐. Best example of role decomposition for complex tasks.

4. SuperAGI (SuperAGI-Org/SuperAGI)

Full platform for building, managing, and running autonomous agents: tool marketplace, memory, telemetry, GUI. ~12k ⭐. Production-ready framework.

5. AgentGPT (reworkd/AgentGPT)

Web UI to assemble and deploy autonomous agents in the browser: name it, set a goal, it generates tasks and executes. ~30k ⭐. Makes agents accessible beyond developers.

How to update this list

Data in this project lives in js/data.js — ask the agent to update the post, it will check current stars via the GitHub API and fix the text.