# openai/multi-agent-emergence-environments

Environment generation code for the paper "Emergent Tool Use From Multi-Agent Autocurricula"

Repository: https://github.com/openai/multi-agent-emergence-environments
Canonical: https://ross.abutalabs.com/products/multi-agent-emergence-environments
Language: Python
License: MIT
License Family: permissive
Archived: true
Last push: 2024-07-30T20:54:31+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2578, "days_push": 764, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1814, forks 325 (observed 2026-08-28T04:05:39.820993+00:00)

## What it is
OpenAI's environment generation code for multi-agent reinforcement learning research, implementing environments like Hide and Seek from the paper 'Emergent Tool Use From Multi-Agent Autocurricula'. It is built on MuJoCo and mujoco-worldgen, providing composable EnvModules and Gym wrappers for constructing multi-agent physics-based games.

## Use cases
- reproduce the hide and seek multi-agent experiments from the OpenAI paper
- build custom multi-agent physics environments with boxes, ramps, and walls
- study emergent tool use and autocurricula in multi-agent RL
- run saved policies in hide and seek environments for analysis
- construct transfer tasks like lock and return or blueprint construction

## When to choose
- you need the exact environments from the Emergent Tool Use paper for reproduction or follow-up research
- you want a modular MuJoCo-based framework for composing multi-agent environments
- you are studying multi-agent autocurricula and emergent behaviors

## When to avoid
- you need maintained software with active support or bug fixes
- you want modern Python versions beyond 3.6 or Windows support
- you need lightweight or non-physics multi-agent environments
- you are looking for a production-ready RL training framework rather than research environments

## Facets
- artifact type: library
- maturity: abandoned
- function: simulation, machine-learning, reinforcement-learning
- domain: reinforcement-learning, artificial-intelligence, simulation
- platform: python
- tags: multi-agent, mujoco, gym-environments, emergent-behavior, research-code, archived, research, linux, macos

## Member repositories
- openai/multi-agent-emergence-environments (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:39.820993+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:21:06.164001+00:00, confidence not recorded.
  - readme: https://github.com/openai/multi-agent-emergence-environments (fetched 2026-08-28T04:05:39.820993+00:00, sha 6d45e4e1d435)
- Data as of 2026-08-30T08:39:29.467469+00:00.