Omnigent: Open Source Meta-Harness for EVERY Coding Agent - Claude Code, Codex, Pi, More
🤖 Omnigent is a free, open-source "meta-harness" from the Databricks AI team that sits on top of the AI agents you already use, like Claude Code and Codex, so you can combine them, control them, and share live sessions across your terminal, web UI, desktop app, and phone. 🚀 In this deep dive I take you from zero to a full multi-agent setup: local install, the web UI, built-in orchestrators, policies, forking, a custom PR review agent, deploying the server to Railway for your team, and running it on Databricks (including the free-edition limits you should know). In this video, you'll learn: ✅ What a meta-harness is, and Omnigent's server vs host (runner) architecture ✅ Installing locally: Python 3.12+, Node.js, and tmux ✅ Setup that auto-detects your configured agents (Claude, Codex) ✅ A full tour of the Omnigent web UI ✅ The two built-in orchestrators: Debby (splits and debates across models) and Polly (implements with one agent, cross-reviews with another) ✅ Policies and guardrails that are actually enforced ✅ Desktop app and session sync across app and terminal ✅ The server vs host concept, and why it matters ✅ Git worktrees, forking sessions, and pair-style workflows ✅ Building your own custom PR review agent ✅ Deploying the server to Railway to share and collaborate ✅ Inviting members and setting permissions ✅ Using Omnigent on Databricks, plus free-edition limitations 💸 Cost: Omnigent itself is 100% FREE and open source. You bring your own model access (an Anthropic or OpenAI key or login), Railway has a free tier, and I cover what does and does not work on Databricks Free Edition. ──────────────────────────────────────── C) TIMESTAMPS ──────────────────────────────────────── 00:00 Omnigent Overview 04:23 Official Announcement 05:02 Architecture Diagram 07:11 Docs And Setup 09:02 Install Requirements 09:54 Agent Setup 11:53 Web UI Tour 14:03 Devi Orchestrator 17:59 Policies Guardrails 19:09 Desktop App Sync 22:06 Server Host Concepts 24:39 Collaboration Comments 26:09 Polly Worktrees Flow 28:37 Cross Review Workflow 29:38 Managing Agents Sessions 30:27 Parallel Architecture Analysis 31:58 Build Custom Agent 34:50 Fork Sessions Pairing 36:03 Deploy Server Railway 42:47 Sharing Inviting Members 46:45 Databricks Setup Limits 54:28 Wrap Up ──────────────────────────────────────── D) LINKS & RESOURCES ──────────────────────────────────────── ↗ Omnigent site: https://omnigent.ai ↗ Install guide: https://omnigent.ai/quickstart/install ↗ Contextual Policies: https://omnigent.ai/docs/policies/overview ↗ GitHub repo: https://github.com/omnigent-ai/omnigent ↗ v0.3.0 release notes: https://github.com/omnigent-ai/omnigent/releases/tag/v0.3.0 ↗ Discord community: https://discord.gg/omnigent Databricks: ↗ Omnigent on Databricks: https://docs.databricks.com/aws/en/omnigent/ ↗ Quickstart (Databricks): https://docs.databricks.com/aws/en/omnigent/quickstart ↗ Databricks blog (intro): https://www.databricks.com/blog/introducing-omnigent-meta-harness-combine-control-and-share-your-agents Hosting and model access: ↗ Railway (free tier): https://railway.app ↗ Anthropic API key (for Claude): https://console.anthropic.com ↗ OpenAI API key (for Codex / GPT): https://platform.openai.com 📁 Demo files used in this video: https://github.com/sudarshan-koirala/omnigent-yt.git ───────────────────────────────────────────────────────────────── ☕ Buy me a Coffee: https://ko-fi.com/datasciencebasics ✌️ Patreon: https://www.patreon.com/datasciencebasics ───────────────────────────────────────────────────────────────── 🤝 Connect with me: 📺 Youtube: https://www.youtube.com/@datasciencebasics?sub_confirmation=1 👔 LinkedIn: https://www.linkedin.com/in/sudarshan-koirala/ 🐦 Twitter: https://twitter.com/mesudarshan 🔉 Medium: https://medium.com/@sudarshan-koirala 💼 Consulting: https://topmate.io/sudarshan_koirala #omnigent #llm #databricks #omnigenttutorial #metaharness #harnessengineering