Guides

Practical how-to guides for working with OpenEnv. These guides are task-oriented and help you accomplish specific goals.

🔌 Using Environments

Learn how to connect to and interact with OpenEnv environments.

Concepts →

🛠️ Building Environments

Create your own custom environments for agentic training.

Your First Environment →

🧠 Training

Integrate OpenEnv with RL frameworks for agent training.

RL Integration →

Using Environments

• Concepts - Understand the core environment, action, observation, and reward abstractions

• Auto-Discovery (AutoEnv) - Automatically discover and load environments

• Connecting to Servers - Connect via HTTP, Docker, or Hugging Face Spaces

• Async vs Sync Usage - When and how to use async vs sync clients

Building Environments

• Your First Environment - Build a simple environment from scratch

• Environment Anatomy - Deep dive into environment structure

• Deployment - Deploy to Docker, Hugging Face Spaces, and registries

• Customizing the Web UI - Customize your environment’s built-in web interface

Training

• RL Framework Integration - Use OpenEnv with TRL, torchforge, and more

• Reward Design - Design effective reward functions for your agents

• Simulation vs Production Mode - When to use the training loop vs direct MCP access

• MCP Environment Lifecycle - How MCP tools fit into the step loop