ai-agent

SkyRL: Modular RL for Founders

Name: SkyRL: Modular RL for Founders
Author: NovaSky-AI

Get full-stack reinforcement learning with SkyRL, a modular library for startup founders, with 2.1k+ GitHub stars

2,053 stars376 forksPythonQuality 8/10Updated 7/6/2026100% free · open source

What it does

SkyRL allows startup founders to build and deploy full-stack reinforcement learning models with ease, using a modular library designed for large language models

Install / run

pip install git+https://github.com/NovaSky-AI/SkyRL.git

When to use it

•You need to train an AI agent to interact with a complex environment
•You want to integrate reinforcement learning with your existing large language model
•You need a modular and flexible library to build custom RL models

Quick start

1Clone the SkyRL repository and navigate to the root directory: `git clone https://github.com/NovaSky-AI/SkyRL.git && cd SkyRL`
2Run the example script: `python examples/cartpole.py` to test the library
3Import the SkyRL library in your Python script: `from skyrl import Agent, Env`
4Define your environment and agent: `env = Env(gym.make('CartPole-v1'))` and `agent = Agent(env, policy='dqn')`
5Train your agent: `agent.train(episodes=100)`

Ready-to-paste prompt

python examples/cartpole.py --policy dqn --episodes 100 --render

Heads up: Make sure you have the `gym` library installed, as it is required by SkyRL for environment simulations, you can install it by running `pip install gym`

Saves to your device

What's inside — free to inspect

No purchase needed

Read the entire source before you build — unlike paid marketplaces that hide it behind a buy button.

top-level files

folders

35.5M

repo size

Apache-2.0

license

Key files

.pre-commit-config.yaml