OKBrain Harness is designed so you can connect any AI model of your choice. You do not need to modify the system for each model — it adapts as needed.
Beyond that, OKBrain Harness supports multi-agent conversations. Think of it as a group chat where each model sees the others' messages, has self-awareness, and knows about the other assistants.
<img src="/images/okbrain-multi-agent.webp" alt="Multi-agent conversation in OKBrain">
By default, we use Google Gemini models because it is easy to get an API key (assuming you have a Google account). Gemini 3 Flash is a very capable model at an affordable price point, and it is fast.
We also have built-in providers for the following — you only need an API key to set them up:
<ul>
<li>Anthropic</li>
<li>xAI</li>
<li>OpenRouter</li>
<li>Fireworks</li>
<li><a href="http://Z.ai">Z.ai</a> (GLM-based models)</li>
<li>Local Ollama models</li>
</ul>
If your preferred model provider is not listed, just ask a coding agent to add support for it. It will take a few minutes, and the included skills will keep everything aligned.

Multi-Agent Conversations

Introduction

Quick Start Guide

Proper Dev Setup

Own & Keep Your Data

Infinite Memory

Internet/Web Use

Computer Use

Events/Calendar Management

Documents Support

Apps with a Coding Agent

Users & Sharing

PWA & Mobile App

Using Local Models

Location Data

Enabling Additional Features

Production Deployment to a Cloud VM

Production Deployment to a Local Mac

Contributions