Skip to content

02 - Open Office

Problem Existing RL setups lack the organizational complexity, role-based coordination, and dynamic market conditions that define how real companies operate.

Code Demo Video Cerebral Valley

Core trick

Models a startup's go-to-market roles as RL agents that each receive a local observation and a reward signal from a customer oracle agent.

Tags

agents

Description

Problem Existing RL setups lack the organizational complexity, role-based coordination, and dynamic market conditions that define how real companies operate. There is no adequate environment for training agents to collaborate and compete under real strategic pressure. Solution A multi-agent RL environment modeled on a startup's go-to-market motion. Agents represent real roles: CEO, Marketing, Sales, Dev, Planning, Content, and a Customer reward oracle. Each observes its local state, takes actions, receives a reward signal, and learns optimal behavior across episodes while coordinating with the full system. Five scenarios stress-test the environment: Baseline GTM Launch, Competitor Launch, Series A Pressure, Churn Spike, and Viral Moment. Multi-agent. Multi-mind. One office.
bvsbharat/OpenOfficeRL★ 215 forksPythonApache-2.0Created Mar 7, 2026 · last push Mar 16, 2026

README excerpt

<div align="center">

# OpenOffice RL

### Multi-Agent Reinforcement Learning Environment for Organizational Decision-Making

**Multi-agent. Multi-mind. One office.**

[![Built on OpenEnv](https://img.shields.io/badge/Built%20on-OpenEnv-blue)](https://github.com/meta-pytorch/OpenEnv)
[![Python 3.10+](https://img.shields.io/badge/Python-3.10%2B-green)](https://python.org)
[![React + Phaser + Three.js](https://img.shields.io/badge/Frontend-React%20%2B%20Phaser%20%2B%20Three.js-61DAFB)](https://reactjs.org)
[![License: Apache 2.0](https://img.shields.io/badge/License-Apache%202.0-orange)](LICENSE)
[![Trained LoRA Adapters](https://img.shields.io/badge/HuggingFace-LoRA%20Adapters-yellow)](https://huggingface.co/HarshalH/office-os-loras)
[![Demo Video](https://img.shields.io/badge/Demo-YouTube-red)](https://youtu.be/dfWrsrGQCQo)

[**Watch the Demo Video**](https://youtu.be/dfWrsrGQCQo)

</div>

---

## What Is This?

A **Gymnasium-compatible multi-agent RL environment** where seven LLM-powered agents run a startup together. The "world" is a startup office; the agents are its people: CEO, Dev, Marketing, Sales, Content, HR, and a Customer oracle. Each agent observes local state, takes role-scoped actions, receives asymmetric reward signals, and coordinates with every other agent through a shared memory board.

Five adversarial scenarios stress-test the environment: **Baseline GTM Launch**, **Competitor Launch**, **Series A Pressure**, **Churn Spike**, and **Viral Moment** — each sh

Team