Welcome to the Arena Documentation

Arena is AgileRL’s hosted platform for training reinforcement learning agents and putting them into production. You organize work in projects, configure experiments in a six-step wizard, run training on cloud or on-prem resource classes, then deploy agents for inference.

Home shows tutorials, recent work, and platform updates. The sidebar links to environments, datasets, experiments, and agents.

Where to start

If you want to…	Start here
Run your first training job	Quickstart
Follow step-by-step tutorials with screenshots	Tutorials
Understand classic RL vs LLM/advanced projects	Classic vs advanced
Manage billing and credits	Credits and plans
Use Arena from the terminal	Arena CLI

Main areas

• Environments: built-in gym sims and your own Python environments

• Datasets: reasoning, preference, SFT, tabular, and non-tabular data for advanced training

• Experiments: wizard, training runs, halt and resume

• Agents: deploy checkpoints and run inference from the UI or terminal

• Account: profile, CLI keys, organizations, usage

Reference

• Glossary

• Algorithms

• Experiment statuses

• Plans and permissions

• Finding your way

Something broken? See Troubleshooting.