AEnvironment
A standardized environment platform for agentic AI development and reinforcement learning, treating benchmarks, tools and even other agents as a single, swappable kind of environment.
AEnvironment is a unified environment platform for agentic reinforcement learning, built around treating everything, from a simple tool function to a complex multi-agent system, as an environment behind one interface.
It ships built-in, ready-to-use benchmark environments, currently TAU2-Bench, SWE-Bench and Terminal-Bench, with native Model Context Protocol support and OpenAI Agents SDK compatibility so a developer can focus on agent logic rather than environment setup. Agents themselves can be wrapped as environments, enabling multi-agent orchestration compatible with mainstream agent frameworks.
A CLI manages instances and services for deployment, and a Claude Code skill automates deployment across local builds, existing images and registered environments. Within Ant Group it is a core environment layer integrated with the AReaL reinforcement learning framework for large-scale agentic RL training.