omnigent-ai/omnigent

Omnigent is an open-source AI agent framework that orchestrates multiple coding agents—Claude Code, Codex, Cursor, Pi, and custom agents—under a unified control layer.

The problem Omnigent solves is the fragmentation of agent tooling: each agent platform has its own interface, policies, and execution model, forcing developers to learn and manage multiple systems separately. Omnigent provides a meta-harness that abstracts these differences away. You define agents once in YAML or code, then swap between different agent backends without rewriting your orchestration logic. The framework enforces policies and sandboxing at the harness level, so governance rules apply consistently regardless of which agent you're using. Sessions persist across devices—start work in your terminal, continue in a browser, pick it up on your phone—with messages, files, and agent state staying synchronized.

Omnigent suits teams that want to experiment with multiple agent platforms without vendor lock-in, or that need to run agents in cloud sandboxes for isolation and scalability. It's particularly valuable if you need real-time collaboration, where teammates can watch an agent work live, co-drive sessions, or fork conversations. The tool supports a wide range of sandbox providers including Modal, Kubernetes, E2B, and Databricks, so you can choose your execution environment independently of your agent choice. If your workflow is tied to a single agent platform, the abstraction layer may add unnecessary complexity.

Almost all open issues are raised by outside users rather than the core team, indicating a substantial base of adopters reporting real-world use. Maintainers typically respond to new issues and pull requests within hours. Work in the issue tracker is dominated by triaged items, medium-priority bugs, and bug fixes.