Uploaded August 2026 | Updated September 2026, 2 weeks ago
Omnigent is an open-source AI agent framework and meta-harness built by the Databricks AI team and Neon. It provides one orchestration layer for Claude Code, Codex, Cursor, Pi, and custom agents, allowing developers to combine or switch between agent harnesses without rebuilding their workflows.
Here is how the architecture works:
🏗️ The Runner wraps each agent in a sandboxed session with a uniform interface, regardless of which model or harness it uses.
🛡️ The Server acts as the control plane, maintaining shared session history and enforcing contextual policies for cost limits, tool permissions, security rules, and human approvals.
📱 The same live agent session can be accessed through the terminal, browser, desktop app, mobile device, or REST API.
In this video, Olga Romanova, FDE, demonstrates Debby, Omnigent’s built-in multi-model debate agent. Claude and GPT independently evaluate when AGI might arrive, then critique each other’s reasoning through multiple rounds before refining their conclusions.
Explore the project and start contributing: omnigent.ai
Omnigent is an open-source AI agent framework and meta-harness built by the Databricks AI team and Neon. It provides one orchestration layer for Claude Code, Codex, Cursor, Pi, and custom agents, allowing developers to combine or switch between agent harnesses without rebuilding their workflows.
Here is how the architecture works:
🏗️ The Runner wraps each agent in a sandboxed session with a uniform interface, regardless of which model or harness it uses.
🛡️ The Server acts as the control plane, maintaining shared session history and enforcing contextual policies for cost limits, tool permissions, security rules, and human approvals.
📱 The same live agent session can be accessed through the terminal, browser, desktop app, mobile device, or REST API.
In this video, Olga Romanova, FDE, demonstrates Debby, Omnigent’s built-in multi-model debate agent. Claude and GPT independently evaluate when AGI might arrive, then critique each other’s reasoning through multiple rounds before refining their conclusions.
Explore the project and start contributing: omnigent.ai
