Descriptions:
The way the best engineers use AI is shifting fast. Instead of committing to a single provider, they use the right model for each job – like using Claude for implementation, and Codex for reviewing, and Kimi/Minimax for simple tasks. The piece we have been missing is a clean way to orchestrate all of those models together. That is exactly what Omnigent is: a meta-harness that runs Claude Code, Codex, and Pi as interchangeable workers under one roof!
In this video I show you how to get it running and put it to work live. One AI agent plans a coding task, hands the implementation to Claude Code in its own git worktree, then routes the finished diff to a different provider, Codex, for review. You also get a governance layer for human in the loop, and sessions that follow you from terminal to browser to phone. And since I build my own agent engine, Archon, every day, I will give you a straight take on where a meta-harness fits.
~~~~~~~~~~~~~~~~~~~~~~~~~~
– Omnigent is open source, try it out now! It’s so easy to get set up:
https://github.com/omnigent-ai/omnigent
~~~~~~~~~~~~~~~~~~~~~~~~~~
– The Dynamous Agentic Coding Course is now FULLY released – learn how to build reliable and repeatable systems for AI coding:
https://dynamous.ai/agentic-coding-course
~~~~~~~~~~~~~~~~~~~~~~~~~~
0:00 Introduction
1:23 Why Meta-Harnesses Matter
3:19 Getting Started
5:51 Polly: Claude Implements, Codex Reviews
7:56 Anatomy of an Orchestrator
10:07 Custom Agents & Guardrails
11:54 Debby: Multi-Model Debate
13:11 Same Session Across Devices
14:18 Wrap-Up
~~~~~~~~~~~~~~~~~~~~~~~~~~
Join me as I push the limits of what is possible with AI. I’ll be uploading videos weekly – at least every Wednesday at 7:00 PM CDT!
