Databricks and Neon Launch Omnigent: A Unified Platform for Managing Multiple AI Agents
Key Takeaways
- ▸Omnigent unifies multiple AI agent frameworks (Claude Code, Codex, Pi, custom agents) in a single interface, allowing seamless swapping and combination without code rewrites
- ▸The platform enables real-time team collaboration and cross-device agent sessions while maintaining fine-grained governance through policies, spend caps, and tool restrictions
- ▸Cloud sandbox support via Modal and Daytona allows agents to run in disposable environments without requiring local hardware, accessible from CLI or server provisioning
Summary
Databricks and Neon have announced Omnigent, an open-source meta-harness platform that provides a unified interface for managing and orchestrating multiple AI agents, including Claude Code, Codex, Pi, and custom YAML-defined agents. The platform enables users to swap or combine different agent harnesses without rewriting code, while maintaining governance through policies and sandboxing.
Omnigent supports cross-device collaboration, allowing users to start agent sessions in the terminal, continue in the browser, and sync to mobile devices. Teams can share sessions in real-time, with agents running in disposable cloud sandboxes via Modal or Daytona—eliminating the need for local hardware. Users maintain full control over agent behavior through spend caps, approval gates for risky actions, and tool-level restrictions applied at the server, agent, or chat level.
The platform integrates with any model provider, including first-party API keys, Claude/ChatGPT subscriptions, and compatible gateways. Installation is available via a single command, with macOS desktop app support and a local web UI. The project is fully open-source and hosted on GitHub, with example agents (Polly and Debby) included to help users get started.
- Open-source architecture supports any model provider—API keys, subscriptions, or compatible gateways—making it agnostic to underlying LLM infrastructure
Editorial Opinion
Omnigent addresses a critical gap in AI agent management: the fragmentation of different agent frameworks and the lack of robust governance at scale. By providing a unified control plane with real-time collaboration, cross-device mobility, and strong policy enforcement, the platform is well-positioned to become essential infrastructure for teams scaling AI agent usage. The open-source approach and support for multiple harnesses signals confidence in the product's flexibility and lowers switching costs.
