agentenv-framework

Creating realistic RL environments requires collaboration between researchers, engineers, and domain experts across many dimensions: artifacts, environments too

What is agentenv-framework?

Creating realistic RL environments requires collaboration between researchers, engineers, and domain experts across many dimensions: artifacts, environments tools, dynamism of the environment, reproducibility, and more. There is no open source framework for building these environments effectively. Until now.

Overview of agentenv-framework

What it is

AgentEnv Framework is a Python SDK and CLI for building, deploying, and running agentic environments and the tasks that grade agents inside them. Environments are containerized servers that speak the open agentenv-framework-protocol. The framework builds them into versioned images, deploys them behind a gateway, points an agent at them, and scores what the agent did.Documentation is available at www.agentenvframework.com/docs.

Key features

  • Builds, deploys, and runs agentic environments as containerized servers using an open protocol.
  • Provides a CLI (agent-env) and SDK for interacting with environments, tasks, and agents.
  • Supports versioned environment images, a registry, and plugins that can add environment types, providers, task steps, and stores.
  • Offers cloud store backends as extras: AWS (S3, Secrets Manager, DynamoDB, ECR) and GCP (Cloud Storage, Secret Manager, Firestore, Artifact Registry).
  • Includes timezone support via a gateway tool (get_time) and configurable task steps.
  • Automates releases with version bumps in pyproject.toml and vX.Y.Z tags; both the framework and protocol are published to PyPI.

Use cases

Use AgentEnv Framework to build, deploy, and score agentic environments and their grading tasks. It is suited for researchers, engineers, and domain experts creating realistic reinforcement learning environments that require collaboration across multiple dimensions such as artifacts, environment tools, dynamism, and reproducibility.

Getting started

Requires Python 3.11.4 or newer and, for local runs, a running Docker daemon. Install the agent-env command as a tool with uv:bashuv tool install agentenv-frameworkagent-env run helloOr run with uvx without installing:bashuvx --from agentenv-framework agent-env run helloOr install with pip in a virtual environment:bashpip install agentenv-frameworkTo use the SDK in your project, add it as a dependency (e.g., uv add agentenv-framework). The cloud store backends require extra installs (e.g., pip install 'agentenv-framework[aws]').

License and pricing

The framework is licensed under the Apache License 2.0, as stated in the LICENSE and NOTICE files. Additional third-party dependency licenses are listed in THIRD_PARTY_NOTICES.md. Pricing is not stated.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/scaleapiagentenv-framework.svg)](https://yunodb.com/tools/scaleapiagentenv-framework?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/scaleapiagentenv-framework?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/scaleapiagentenv-framework.svg" alt="Featured on YunoDB" height="20"></a>