openai/universe-starter-agent resource

Health v2 · maintenance only

Flags: no_releases archived

Adoption not part of the score

1097 stars · 313 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A starter implementation of an A3C reinforcement learning agent that can solve OpenAI Universe environments, including Atari games like Pong. It is a reference codebase for learning policy-gradient RL, now deprecated in favor of OpenAI's Retro library.

Use cases

• train an A3C agent on Atari Pong
• learn how asynchronous policy gradient RL is implemented
• get started with OpenAI Universe environments
• run parallel RL workers with a parameter server
• study a real-time RL agent codebase in TensorFlow

When to choose

• you want a minimal, readable A3C reference implementation
• you are learning reinforcement learning with TensorFlow and OpenAI Gym
• you need a starting point for agents on Universe/VNC environments

When to avoid

• you need maintained software - it is deprecated in favor of OpenAI Retro
• you want modern TensorFlow or PyTorch versions
• you need production RL training infrastructure

Facets

learning-resource · maturity abandoned

reinforcement-learning machine-learning agent-framework benchmarking reinforcement-learning machine-learning artificial-intelligence gaming-tools tutorials python a3c openai-gym universe-environments atari starter-agent deprecated tensorflow linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="openai/universe-starter-agent")