---
id: openenv-core
version: "0.3.0"
license: unclear
license_treatment: unclear
maintenance: active
---

# openenv-core — A unified framework for reinforcement learning environments

License: unclear · Maintenance: active · Downloads: 452.8K/mo

## What it is and what it does

OpenEnv is a framework for building and deploying isolated execution environments that agents can train against using reinforcement learning. It standardizes the agent-environment interaction using Gymnasium-style APIs (`reset`, `step`, `state`) and handles the networking layer via WebSocket and HTTP, allowing agents to interact with environments running in Docker containers or on remote services.

The package provides both server-side components (base classes for implementing environments, FastAPI integration, web UI) and client-side components (async/sync wrappers for connecting to environments, type-safe action/observation handling). It is built on top of fastapi, pydantic, uvicorn, and other production frameworks, and integrates with huggingface_hub and openai. The framework is currently experimental and undergoing active development.

Use it for:

- Train agents to play games or solve tasks by wrapping task logic in an OpenEnv environment and connecting agents via the client API.
- Deploy custom agent execution sandboxes as isolated Docker containers accessible over HTTP, with automatic tool discovery and action routing.
- Build interactive web-based debugging interfaces for environment behavior using the built-in web UI and real-time WebSocket updates.
- Integrate RL training loops with fastapi and uvicorn by using the async client to step environments and collect rewards during training.
- Share reproducible environments across teams or platforms by packaging them as Docker images and hosting on remote services.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

OpenEnv provides a Gymnasium-style framework for creating and deploying isolated execution environments that agents can interact with via standard `reset()`, `step()`, and `state()` APIs over HTTP and WebSocket.

Yes, if you are actively building or training agents against custom environments and can tolerate experimental APIs. The framework is actively maintained and has low install friction, but it is explicitly in early development with expected breaking changes. Verify the license before use. Not recommended for production systems requiring API stability.

## Install

pip install openenv-core
uv add openenv-core
poetry add openenv-core

## Installing openenv-core

Before you install: Low install friction with a pure-Python wheel. Active maintenance (latest release 95 days ago). Requires Python 3.10+. The 15 runtime dependencies include heavy frameworks (fastapi, pydantic, uvicorn, huggingface_hub, openai, gradio) that will pull in substantial transitive closure.

License in practice: License treatment is unclear—no SPDX identifier or raw license text in metadata. Verify the actual license before use in proprietary or copyleft-sensitive projects.

Quickstart:

pip install openenv-core

import asyncio
from fastapi import FastAPI
from openenv_core import Environment

async def main():
    # Connect to environment via client
    pass

asyncio.run(main())

Requires Python 3.10+. A running environment server (local Docker or remote service) is needed to connect to. Environment clients must be installed separately.

Verify before relying:

- Stability and API surface—marked as experimental with expected bugs and breaking changes; production readiness unclear.
- Whether the framework supports custom environment implementations or only pre-built clients.
- Performance characteristics and scalability limits for concurrent agents or large action/observation spaces.
- Actual license terms—metadata shows no license, so legal status is unknown.

## Package facts

- License: not declared (unclear)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 452.8K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

agentic RL training environments, gymnasium-style environment framework, isolated execution environments for agents, reinforcement learning environment deployment, agent environment interaction protocol, environment server with websocket, docker-based agent sandboxes, reinforcement-learning, agent-training, environment-framework

[View on SkillFed](https://skillfed.io/packages/openenv-core) · [View on PyPI](https://pypi.org/project/openenv-core/)